191 files changed, 6972 insertions, 5184 deletions

diff --git a/CMakeModules/CopyYuzuQt5Deps.cmake b/CMakeModules/CopyYuzuQt5Deps.cmake
index 1e9810bba..2598b9b60 100644
--- a/CMakeModules/CopyYuzuQt5Deps.cmake
+++ b/CMakeModules/CopyYuzuQt5Deps.cmake
@@ -6,9 +6,9 @@ function(copy_yuzu_Qt5_deps target_dir)
     set(Qt5_STYLES_DIR "${Qt5_DIR}/../../../plugins/styles/")
     set(Qt5_IMAGEFORMATS_DIR "${Qt5_DIR}/../../../plugins/imageformats/")
     set(Qt5_RESOURCES_DIR "${Qt5_DIR}/../../../resources/")
-    set(PLATFORMS ${DLL_DEST}platforms/)
-    set(STYLES ${DLL_DEST}styles/)
-    set(IMAGEFORMATS ${DLL_DEST}imageformats/)
+    set(PLATFORMS ${DLL_DEST}plugins/platforms/)
+    set(STYLES ${DLL_DEST}plugins/styles/)
+    set(IMAGEFORMATS ${DLL_DEST}plugins/imageformats/)
     windows_copy_files(${target_dir} ${Qt5_DLL_DIR} ${DLL_DEST}
         icudt*.dll
         icuin*.dll
@@ -42,11 +42,15 @@ function(copy_yuzu_Qt5_deps target_dir)
             icudtl.dat
         )
     endif ()
-
     windows_copy_files(yuzu ${Qt5_PLATFORMS_DIR} ${PLATFORMS} qwindows$<$<CONFIG:Debug>:d>.*)
     windows_copy_files(yuzu ${Qt5_STYLES_DIR} ${STYLES} qwindowsvistastyle$<$<CONFIG:Debug>:d>.*)
     windows_copy_files(yuzu ${Qt5_IMAGEFORMATS_DIR} ${IMAGEFORMATS}
         qjpeg$<$<CONFIG:Debug>:d>.*
         qgif$<$<CONFIG:Debug>:d>.*
         )
+    # Create an empty qt.conf file. Qt will detect that this file exists, and use the folder that its in as the root folder.
+    # This way it'll look for plugins in the root/plugins/ folder
+    add_custom_command(TARGET yuzu POST_BUILD
+        COMMAND ${CMAKE_COMMAND} -E touch ${DLL_DEST}qt.conf
+    )
 endfunction(copy_yuzu_Qt5_deps)
diff --git a/CMakeModules/GenerateSCMRev.cmake b/CMakeModules/GenerateSCMRev.cmake
index fa7ae835f..83e4e9df2 100644
--- a/CMakeModules/GenerateSCMRev.cmake
+++ b/CMakeModules/GenerateSCMRev.cmake
@@ -57,8 +57,6 @@ set(HASH_FILES
     "${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"
     "${VIDEO_CORE}/shader/decode/arithmetic_half_immediate.cpp"
@@ -91,8 +89,6 @@ set(HASH_FILES
     "${VIDEO_CORE}/shader/ast.h"
     "${VIDEO_CORE}/shader/compiler_settings.cpp"
     "${VIDEO_CORE}/shader/compiler_settings.h"
-    "${VIDEO_CORE}/shader/const_buffer_locker.cpp"
-    "${VIDEO_CORE}/shader/const_buffer_locker.h"
     "${VIDEO_CORE}/shader/control_flow.cpp"
     "${VIDEO_CORE}/shader/control_flow.h"
     "${VIDEO_CORE}/shader/decode.cpp"
@@ -101,9 +97,13 @@ set(HASH_FILES
     "${VIDEO_CORE}/shader/node.h"
     "${VIDEO_CORE}/shader/node_helper.cpp"
     "${VIDEO_CORE}/shader/node_helper.h"
+    "${VIDEO_CORE}/shader/registry.cpp"
+    "${VIDEO_CORE}/shader/registry.h"
     "${VIDEO_CORE}/shader/shader_ir.cpp"
     "${VIDEO_CORE}/shader/shader_ir.h"
     "${VIDEO_CORE}/shader/track.cpp"
+    "${VIDEO_CORE}/shader/transform_feedback.cpp"
+    "${VIDEO_CORE}/shader/transform_feedback.h"
 )
 set(COMBINED "")
 foreach (F IN LISTS HASH_FILES)
diff --git a/README.md b/README.md
index bdee2e872..1e9f67e08 100644
--- a/README.md
+++ b/README.md
@@ -1,7 +1,8 @@
 yuzu emulator
 =============
-[![Travis CI Build Status](https://travis-ci.org/yuzu-emu/yuzu.svg?branch=master)](https://travis-ci.org/yuzu-emu/yuzu)
+[![Travis CI Build Status](https://travis-ci.com/yuzu-emu/yuzu.svg?branch=master)](https://travis-ci.com/yuzu-emu/yuzu)
 [![Azure Mainline CI Build Status](https://dev.azure.com/yuzu-emu/yuzu/_apis/build/status/yuzu%20mainline?branchName=master)](https://dev.azure.com/yuzu-emu/yuzu/)
+[![Discord](https://img.shields.io/discord/398318088170242053?color=%237289DA&label=yuzu&logo=discord&logoColor=white)](https://discord.gg/XQV6dn9)
 
 yuzu is an experimental open-source emulator for the Nintendo Switch from the creators of [Citra](https://citra-emu.org/).
 
@@ -21,7 +22,7 @@ For development discussion, please join us on [Discord](https://discord.gg/XQV6d
 
 Most of the development happens on GitHub. It's also where [our central repository](https://github.com/yuzu-emu/yuzu) is hosted.
 
-If you want to contribute please take a look at the [Contributor's Guide](CONTRIBUTING.md) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should as well contact any of the developers on Discord in order to know about the current state of the emulator.
+If you want to contribute please take a look at the [Contributor's Guide](https://github.com/yuzu-emu/yuzu/wiki/Contributing) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should also contact any of the developers on Discord in order to know about the current state of the emulator.
 
 ### Building
 
diff --git a/externals/microprofile/microprofile.h b/externals/microprofile/microprofile.h
index cdb312b87..9d830f7bf 100644
--- a/externals/microprofile/microprofile.h
+++ b/externals/microprofile/microprofile.h
@@ -243,6 +243,7 @@ typedef uint32_t ThreadIdType;
 #define MICROPROFILE_DEFINE_GPU(var, name, color) MicroProfileToken g_mp_##var = MicroProfileGetToken("GPU", name, color, MicroProfileTokenTypeGpu)
 #define MICROPROFILE_TOKEN_PASTE0(a, b) a ## b
 #define MICROPROFILE_TOKEN_PASTE(a, b)  MICROPROFILE_TOKEN_PASTE0(a,b)
+#define MICROPROFILE_TOKEN(var) g_mp_##var
 #define MICROPROFILE_SCOPE(var) MicroProfileScopeHandler MICROPROFILE_TOKEN_PASTE(foo, __LINE__)(g_mp_##var)
 #define MICROPROFILE_SCOPE_TOKEN(token) MicroProfileScopeHandler MICROPROFILE_TOKEN_PASTE(foo, __LINE__)(token)
 #define MICROPROFILE_SCOPEI(group, name, color) static MicroProfileToken MICROPROFILE_TOKEN_PASTE(g_mp,__LINE__) = MicroProfileGetToken(group, name, color, MicroProfileTokenTypeCpu); MicroProfileScopeHandler MICROPROFILE_TOKEN_PASTE(foo,__LINE__)( MICROPROFILE_TOKEN_PASTE(g_mp,__LINE__))
@@ -827,7 +828,7 @@ inline MicroProfileLogEntry MicroProfileMakeLogIndex(uint64_t nBegin, MicroProfi
     MicroProfileLogEntry Entry =  (nBegin<<62) | ((0x3fff&nToken)<<48) | (MP_LOG_TICK_MASK&nTick);
     int t = MicroProfileLogType(Entry);
     uint64_t nTimerIndex = MicroProfileLogTimerIndex(Entry);
-    MP_ASSERT(t == nBegin);
+    MP_ASSERT((uint64_t)t == nBegin);
     MP_ASSERT(nTimerIndex == (nToken&0x3fff));
     return Entry;
 
@@ -1555,10 +1556,10 @@ void MicroProfileFlip()
 
         pFramePut->nFrameStartCpu = MP_TICK();
         pFramePut->nFrameStartGpu = (uint32_t)MicroProfileGpuInsertTimeStamp();
-        if(pFrameNext->nFrameStartGpu != (uint64_t)-1)
+        if(pFrameNext->nFrameStartGpu != -1)
             pFrameNext->nFrameStartGpu = MicroProfileGpuGetTimeStamp((uint32_t)pFrameNext->nFrameStartGpu);
 
-        if(pFrameCurrent->nFrameStartGpu == (uint64_t)-1)
+        if(pFrameCurrent->nFrameStartGpu == -1)
             pFrameCurrent->nFrameStartGpu = pFrameNext->nFrameStartGpu + 1;
 
         uint64_t nFrameStartCpu = pFrameCurrent->nFrameStartCpu;
diff --git a/src/audio_core/algorithm/interpolate.cpp b/src/audio_core/algorithm/interpolate.cpp
index a58f24169..49ab9d3e1 100644
--- a/src/audio_core/algorithm/interpolate.cpp
+++ b/src/audio_core/algorithm/interpolate.cpp
@@ -8,13 +8,14 @@
 #include <climits>
 #include <cmath>
 #include <vector>
+
 #include "audio_core/algorithm/interpolate.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 
 namespace AudioCore {
 
-constexpr std::array<s16, 512> curve_lut0 = {
+constexpr std::array<s16, 512> curve_lut0{
     6600,  19426, 6722,  3,     6479,  19424, 6845,  9,     6359,  19419, 6968,  15,    6239,
     19412, 7093,  22,    6121,  19403, 7219,  28,    6004,  19391, 7345,  34,    5888,  19377,
     7472,  41,    5773,  19361, 7600,  48,    5659,  19342, 7728,  55,    5546,  19321, 7857,
@@ -56,7 +57,7 @@ constexpr std::array<s16, 512> curve_lut0 = {
     19403, 6121,  22,    7093,  19412, 6239,  15,    6968,  19419, 6359,  9,     6845,  19424,
     6479,  3,     6722,  19426, 6600};
 
-constexpr std::array<s16, 512> curve_lut1 = {
+constexpr std::array<s16, 512> curve_lut1{
     -68,   32639, 69,    -5,    -200,  32630, 212,   -15,   -328,  32613, 359,   -26,   -450,
     32586, 512,   -36,   -568,  32551, 669,   -47,   -680,  32507, 832,   -58,   -788,  32454,
     1000,  -69,   -891,  32393, 1174,  -80,   -990,  32323, 1352,  -92,   -1084, 32244, 1536,
@@ -98,7 +99,7 @@ constexpr std::array<s16, 512> curve_lut1 = {
     32551, -568,  -36,   512,   32586, -450,  -26,   359,   32613, -328,  -15,   212,   32630,
     -200,  -5,    69,    32639, -68};
 
-constexpr std::array<s16, 512> curve_lut2 = {
+constexpr std::array<s16, 512> curve_lut2{
     3195,  26287, 3329,  -32,   3064,  26281, 3467,  -34,   2936,  26270, 3608,  -38,   2811,
     26253, 3751,  -42,   2688,  26230, 3897,  -46,   2568,  26202, 4046,  -50,   2451,  26169,
     4199,  -54,   2338,  26130, 4354,  -58,   2227,  26085, 4512,  -63,   2120,  26035, 4673,
@@ -146,10 +147,10 @@ std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input,
 
     if (ratio <= 0) {
         LOG_CRITICAL(Audio, "Nonsensical interpolation ratio {}", ratio);
-        ratio = 1.0;
+        return input;
     }
 
-    const int step = static_cast<int>(ratio * 0x8000);
+    const s32 step{static_cast<s32>(ratio * 0x8000)};
     const std::array<s16, 512>& lut = [step] {
         if (step > 0xaaaa) {
             return curve_lut0;
@@ -160,28 +161,37 @@ std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input,
         return curve_lut2;
     }();
 
-    std::vector<s16> output(static_cast<std::size_t>(input.size() / ratio));
-    int in_offset = 0;
-    for (std::size_t out_offset = 0; out_offset < output.size(); out_offset += 2) {
-        const int lut_index = (state.fraction >> 8) * 4;
+    const std::size_t num_frames{input.size() / 2};
+
+    std::vector<s16> output;
+    output.reserve(static_cast<std::size_t>(input.size() / ratio + InterpolationState::taps));
 
-        const int l = input[(in_offset + 0) * 2 + 0] * lut[lut_index + 0] +
-                      input[(in_offset + 1) * 2 + 0] * lut[lut_index + 1] +
-                      input[(in_offset + 2) * 2 + 0] * lut[lut_index + 2] +
-                      input[(in_offset + 3) * 2 + 0] * lut[lut_index + 3];
+    for (std::size_t frame{}; frame < num_frames; ++frame) {
+        const std::size_t lut_index{(state.fraction >> 8) * InterpolationState::taps};
 
-        const int r = input[(in_offset + 0) * 2 + 1] * lut[lut_index + 0] +
-                      input[(in_offset + 1) * 2 + 1] * lut[lut_index + 1] +
-                      input[(in_offset + 2) * 2 + 1] * lut[lut_index + 2] +
-                      input[(in_offset + 3) * 2 + 1] * lut[lut_index + 3];
+        std::rotate(state.history.begin(), state.history.end() - 1, state.history.end());
+        state.history[0][0] = input[frame * 2 + 0];
+        state.history[0][1] = input[frame * 2 + 1];
 
-        const int new_offset = state.fraction + step;
+        while (state.position <= 1.0) {
+            const s32 left{state.history[0][0] * lut[lut_index + 0] +
+                           state.history[1][0] * lut[lut_index + 1] +
+                           state.history[2][0] * lut[lut_index + 2] +
+                           state.history[3][0] * lut[lut_index + 3]};
+            const s32 right{state.history[0][1] * lut[lut_index + 0] +
+                            state.history[1][1] * lut[lut_index + 1] +
+                            state.history[2][1] * lut[lut_index + 2] +
+                            state.history[3][1] * lut[lut_index + 3]};
+            const s32 new_offset{state.fraction + step};
 
-        in_offset += new_offset >> 15;
-        state.fraction = new_offset & 0x7fff;
+            state.fraction = new_offset & 0x7fff;
 
-        output[out_offset + 0] = static_cast<s16>(std::clamp(l >> 15, SHRT_MIN, SHRT_MAX));
-        output[out_offset + 1] = static_cast<s16>(std::clamp(r >> 15, SHRT_MIN, SHRT_MAX));
+            output.emplace_back(static_cast<s16>(std::clamp(left >> 15, SHRT_MIN, SHRT_MAX)));
+            output.emplace_back(static_cast<s16>(std::clamp(right >> 15, SHRT_MIN, SHRT_MAX)));
+
+            state.position += ratio;
+        }
+        state.position -= 1.0;
     }
 
     return output;
diff --git a/src/audio_core/algorithm/interpolate.h b/src/audio_core/algorithm/interpolate.h
index 1b9831a75..ab1a31754 100644
--- a/src/audio_core/algorithm/interpolate.h
+++ b/src/audio_core/algorithm/interpolate.h
@@ -6,12 +6,17 @@
 
 #include <array>
 #include <vector>
+
 #include "common/common_types.h"
 
 namespace AudioCore {
 
 struct InterpolationState {
-    int fraction = 0;
+    static constexpr std::size_t taps{4};
+    static constexpr std::size_t history_size{taps * 2 - 1};
+    std::array<std::array<s16, 2>, history_size> history{};
+    double position{};
+    s32 fraction{};
 };
 
 /// Interpolates input signal to produce output signal.
diff --git a/src/audio_core/cubeb_sink.cpp b/src/audio_core/cubeb_sink.cpp
index 7047ed9cf..c4e0e30fe 100644
--- a/src/audio_core/cubeb_sink.cpp
+++ b/src/audio_core/cubeb_sink.cpp
@@ -8,6 +8,7 @@
 #include "audio_core/cubeb_sink.h"
 #include "audio_core/stream.h"
 #include "audio_core/time_stretch.h"
+#include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/ring_buffer.h"
 #include "core/settings.h"
@@ -65,12 +66,25 @@ public:
     void EnqueueSamples(u32 source_num_channels, const std::vector<s16>& samples) override {
         if (source_num_channels > num_channels) {
             // Downsample 6 channels to 2
+            ASSERT_MSG(source_num_channels == 6, "Channel count must be 6");
+
             std::vector<s16> buf;
             buf.reserve(samples.size() * num_channels / source_num_channels);
             for (std::size_t i = 0; i < samples.size(); i += source_num_channels) {
-                for (std::size_t ch = 0; ch < num_channels; ch++) {
-                    buf.push_back(samples[i + ch]);
-                }
+                // Downmixing implementation taken from the ATSC standard
+                const s16 left{samples[i + 0]};
+                const s16 right{samples[i + 1]};
+                const s16 center{samples[i + 2]};
+                const s16 surround_left{samples[i + 4]};
+                const s16 surround_right{samples[i + 5]};
+                // Not used in the ATSC reference implementation
+                [[maybe_unused]] const s16 low_frequency_effects { samples[i + 3] };
+
+                constexpr s32 clev{707}; // center mixing level coefficient
+                constexpr s32 slev{707}; // surround mixing level coefficient
+
+                buf.push_back(left + (clev * center / 1000) + (slev * surround_left / 1000));
+                buf.push_back(right + (clev * center / 1000) + (slev * surround_right / 1000));
             }
             queue.Push(buf);
             return;
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 9afc6105d..fbebed715 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -38,8 +38,6 @@ add_custom_command(OUTPUT scm_rev.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"
@@ -72,8 +70,6 @@ add_custom_command(OUTPUT scm_rev.cpp
       "${VIDEO_CORE}/shader/ast.h"
       "${VIDEO_CORE}/shader/compiler_settings.cpp"
       "${VIDEO_CORE}/shader/compiler_settings.h"
-      "${VIDEO_CORE}/shader/const_buffer_locker.cpp"
-      "${VIDEO_CORE}/shader/const_buffer_locker.h"
       "${VIDEO_CORE}/shader/control_flow.cpp"
       "${VIDEO_CORE}/shader/control_flow.h"
       "${VIDEO_CORE}/shader/decode.cpp"
@@ -82,9 +78,13 @@ add_custom_command(OUTPUT scm_rev.cpp
       "${VIDEO_CORE}/shader/node.h"
       "${VIDEO_CORE}/shader/node_helper.cpp"
       "${VIDEO_CORE}/shader/node_helper.h"
+      "${VIDEO_CORE}/shader/registry.cpp"
+      "${VIDEO_CORE}/shader/registry.h"
       "${VIDEO_CORE}/shader/shader_ir.cpp"
       "${VIDEO_CORE}/shader/shader_ir.h"
       "${VIDEO_CORE}/shader/track.cpp"
+      "${VIDEO_CORE}/shader/transform_feedback.cpp"
+      "${VIDEO_CORE}/shader/transform_feedback.h"
       # 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"
diff --git a/src/common/math_util.h b/src/common/math_util.h
index d6c35ee89..83ef0201f 100644
--- a/src/common/math_util.h
+++ b/src/common/math_util.h
@@ -24,17 +24,29 @@ struct Rectangle {
         : left(left), top(top), right(right), bottom(bottom) {}
 
     T GetWidth() const {
-        return std::abs(static_cast<std::make_signed_t<T>>(right - left));
+        if constexpr (std::is_floating_point_v<T>) {
+            return std::abs(right - left);
+        } else {
+            return std::abs(static_cast<std::make_signed_t<T>>(right - left));
+        }
     }
+
     T GetHeight() const {
-        return std::abs(static_cast<std::make_signed_t<T>>(bottom - top));
+        if constexpr (std::is_floating_point_v<T>) {
+            return std::abs(bottom - top);
+        } else {
+            return std::abs(static_cast<std::make_signed_t<T>>(bottom - top));
+        }
     }
+
     Rectangle<T> TranslateX(const T x) const {
         return Rectangle{left + x, top, right + x, bottom};
     }
+
     Rectangle<T> TranslateY(const T y) const {
         return Rectangle{left, top + y, right, bottom + y};
     }
+
     Rectangle<T> Scale(const float s) const {
         return Rectangle{left, top, static_cast<T>(left + GetWidth() * s),
                          static_cast<T>(top + GetHeight() * s)};
diff --git a/src/common/page_table.cpp b/src/common/page_table.cpp
index 69b7abc54..566b57b62 100644
--- a/src/common/page_table.cpp
+++ b/src/common/page_table.cpp
@@ -16,7 +16,6 @@ void PageTable::Resize(std::size_t address_space_width_in_bits) {
 
     pointers.resize(num_page_table_entries);
     attributes.resize(num_page_table_entries);
-    backing_addr.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
@@ -25,6 +24,17 @@ void PageTable::Resize(std::size_t address_space_width_in_bits) {
 
     pointers.shrink_to_fit();
     attributes.shrink_to_fit();
+}
+
+BackingPageTable::BackingPageTable(std::size_t page_size_in_bits) : PageTable{page_size_in_bits} {}
+
+BackingPageTable::~BackingPageTable() = default;
+
+void BackingPageTable::Resize(std::size_t address_space_width_in_bits) {
+    PageTable::Resize(address_space_width_in_bits);
+    const std::size_t num_page_table_entries = 1ULL
+                                               << (address_space_width_in_bits - page_size_in_bits);
+    backing_addr.resize(num_page_table_entries);
     backing_addr.shrink_to_fit();
 }
 
diff --git a/src/common/page_table.h b/src/common/page_table.h
index 8b8ff0bb8..dbc272ab7 100644
--- a/src/common/page_table.h
+++ b/src/common/page_table.h
@@ -76,9 +76,20 @@ struct PageTable {
      */
     std::vector<PageType> attributes;
 
-    std::vector<u64> backing_addr;
-
     const std::size_t page_size_in_bits{};
 };
 
+/**
+ * A more advanced Page Table with the ability to save a backing address when using it
+ * depends on another MMU.
+ */
+struct BackingPageTable : PageTable {
+    explicit BackingPageTable(std::size_t page_size_in_bits);
+    ~BackingPageTable();
+
+    void Resize(std::size_t address_space_width_in_bits);
+
+    std::vector<u64> backing_addr;
+};
+
 } // namespace Common
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index 88c06b2ce..b31a0328c 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -131,8 +131,8 @@ add_library(core STATIC
     frontend/framebuffer_layout.cpp
     frontend/framebuffer_layout.h
     frontend/input.h
-    frontend/scope_acquire_window_context.cpp
-    frontend/scope_acquire_window_context.h
+    frontend/scope_acquire_context.cpp
+    frontend/scope_acquire_context.h
     gdbstub/gdbstub.cpp
     gdbstub/gdbstub.h
     hardware_interrupt_manager.cpp
@@ -595,8 +595,12 @@ endif()
 
 if (ARCHITECTURE_x86_64)
     target_sources(core PRIVATE
-        arm/dynarmic/arm_dynarmic.cpp
-        arm/dynarmic/arm_dynarmic.h
+        arm/dynarmic/arm_dynarmic_32.cpp
+        arm/dynarmic/arm_dynarmic_32.h
+        arm/dynarmic/arm_dynarmic_64.cpp
+        arm/dynarmic/arm_dynarmic_64.h
+        arm/dynarmic/arm_dynarmic_cp15.cpp
+        arm/dynarmic/arm_dynarmic_cp15.h
     )
     target_link_libraries(core PRIVATE dynarmic)
 endif()
diff --git a/src/core/arm/arm_interface.h b/src/core/arm/arm_interface.h
index 47b964eb7..57eae839e 100644
--- a/src/core/arm/arm_interface.h
+++ b/src/core/arm/arm_interface.h
@@ -25,7 +25,20 @@ public:
     explicit ARM_Interface(System& system_) : system{system_} {}
     virtual ~ARM_Interface() = default;
 
-    struct ThreadContext {
+    struct ThreadContext32 {
+        std::array<u32, 16> cpu_registers;
+        u32 cpsr;
+        std::array<u8, 4> padding;
+        std::array<u64, 32> fprs;
+        u32 fpscr;
+        u32 fpexc;
+        u32 tpidr;
+    };
+    // Internally within the kernel, it expects the AArch32 version of the
+    // thread context to be 344 bytes in size.
+    static_assert(sizeof(ThreadContext32) == 0x158);
+
+    struct ThreadContext64 {
         std::array<u64, 31> cpu_registers;
         u64 sp;
         u64 pc;
@@ -38,7 +51,7 @@ public:
     };
     // Internally within the kernel, it expects the AArch64 version of the
     // thread context to be 800 bytes in size.
-    static_assert(sizeof(ThreadContext) == 0x320);
+    static_assert(sizeof(ThreadContext64) == 0x320);
 
     /// Runs the CPU until an event happens
     virtual void Run() = 0;
@@ -130,17 +143,10 @@ public:
      */
     virtual void SetTPIDR_EL0(u64 value) = 0;
 
-    /**
-     * Saves the current CPU context
-     * @param ctx Thread context to save
-     */
-    virtual void SaveContext(ThreadContext& ctx) = 0;
-
-    /**
-     * Loads a CPU context
-     * @param ctx Thread context to load
-     */
-    virtual void LoadContext(const ThreadContext& ctx) = 0;
+    virtual void SaveContext(ThreadContext32& ctx) = 0;
+    virtual void SaveContext(ThreadContext64& ctx) = 0;
+    virtual void LoadContext(const ThreadContext32& ctx) = 0;
+    virtual void LoadContext(const ThreadContext64& ctx) = 0;
 
     /// Clears the exclusive monitor's state.
     virtual void ClearExclusiveState() = 0;
diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.cpp b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
new file mode 100644
index 000000000..187a972ac
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
@@ -0,0 +1,208 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cinttypes>
+#include <memory>
+#include <dynarmic/A32/a32.h>
+#include <dynarmic/A32/config.h>
+#include <dynarmic/A32/context.h>
+#include "common/microprofile.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
+#include "core/arm/dynarmic/arm_dynarmic_cp15.h"
+#include "core/core.h"
+#include "core/core_manager.h"
+#include "core/core_timing.h"
+#include "core/hle/kernel/svc.h"
+#include "core/memory.h"
+
+namespace Core {
+
+class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
+public:
+    explicit DynarmicCallbacks32(ARM_Dynarmic_32& parent) : parent(parent) {}
+
+    u8 MemoryRead8(u32 vaddr) override {
+        return parent.system.Memory().Read8(vaddr);
+    }
+    u16 MemoryRead16(u32 vaddr) override {
+        return parent.system.Memory().Read16(vaddr);
+    }
+    u32 MemoryRead32(u32 vaddr) override {
+        return parent.system.Memory().Read32(vaddr);
+    }
+    u64 MemoryRead64(u32 vaddr) override {
+        return parent.system.Memory().Read64(vaddr);
+    }
+
+    void MemoryWrite8(u32 vaddr, u8 value) override {
+        parent.system.Memory().Write8(vaddr, value);
+    }
+    void MemoryWrite16(u32 vaddr, u16 value) override {
+        parent.system.Memory().Write16(vaddr, value);
+    }
+    void MemoryWrite32(u32 vaddr, u32 value) override {
+        parent.system.Memory().Write32(vaddr, value);
+    }
+    void MemoryWrite64(u32 vaddr, u64 value) override {
+        parent.system.Memory().Write64(vaddr, value);
+    }
+
+    void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
+        UNIMPLEMENTED();
+    }
+
+    void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
+        switch (exception) {
+        case Dynarmic::A32::Exception::UndefinedInstruction:
+        case Dynarmic::A32::Exception::UnpredictableInstruction:
+            break;
+        case Dynarmic::A32::Exception::Breakpoint:
+            break;
+        }
+        LOG_CRITICAL(HW_GPU, "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X})",
+                     static_cast<std::size_t>(exception), pc, MemoryReadCode(pc));
+        UNIMPLEMENTED();
+    }
+
+    void CallSVC(u32 swi) override {
+        Kernel::CallSVC(parent.system, swi);
+    }
+
+    void AddTicks(u64 ticks) override {
+        // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
+        // rough approximation of the amount of executed ticks in the system, it may be thrown off
+        // if not all cores are doing a similar amount of work. Instead of doing this, we should
+        // device a way so that timing is consistent across all cores without increasing the ticks 4
+        // times.
+        u64 amortized_ticks = (ticks - num_interpreted_instructions) / Core::NUM_CPU_CORES;
+        // Always execute at least one tick.
+        amortized_ticks = std::max<u64>(amortized_ticks, 1);
+
+        parent.system.CoreTiming().AddTicks(amortized_ticks);
+        num_interpreted_instructions = 0;
+    }
+    u64 GetTicksRemaining() override {
+        return std::max(parent.system.CoreTiming().GetDowncount(), {});
+    }
+
+    ARM_Dynarmic_32& parent;
+    std::size_t num_interpreted_instructions{};
+    u64 tpidrro_el0{};
+    u64 tpidr_el0{};
+};
+
+std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable& page_table,
+                                                             std::size_t address_space_bits) const {
+    Dynarmic::A32::UserConfig config;
+    config.callbacks = cb.get();
+    // TODO(bunnei): Implement page table for 32-bit
+    // config.page_table = &page_table.pointers;
+    config.coprocessors[15] = std::make_shared<DynarmicCP15>((u32*)&CP15_regs[0]);
+    config.define_unpredictable_behaviour = true;
+    return std::make_unique<Dynarmic::A32::Jit>(config);
+}
+
+MICROPROFILE_DEFINE(ARM_Jit_Dynarmic_32, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
+
+void ARM_Dynarmic_32::Run() {
+    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic_32);
+    jit->Run();
+}
+
+void ARM_Dynarmic_32::Step() {
+    cb->InterpreterFallback(jit->Regs()[15], 1);
+}
+
+ARM_Dynarmic_32::ARM_Dynarmic_32(System& system, ExclusiveMonitor& exclusive_monitor,
+                                 std::size_t core_index)
+    : ARM_Interface{system},
+      cb(std::make_unique<DynarmicCallbacks32>(*this)), core_index{core_index},
+      exclusive_monitor{dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
+
+ARM_Dynarmic_32::~ARM_Dynarmic_32() = default;
+
+void ARM_Dynarmic_32::SetPC(u64 pc) {
+    jit->Regs()[15] = static_cast<u32>(pc);
+}
+
+u64 ARM_Dynarmic_32::GetPC() const {
+    return jit->Regs()[15];
+}
+
+u64 ARM_Dynarmic_32::GetReg(int index) const {
+    return jit->Regs()[index];
+}
+
+void ARM_Dynarmic_32::SetReg(int index, u64 value) {
+    jit->Regs()[index] = static_cast<u32>(value);
+}
+
+u128 ARM_Dynarmic_32::GetVectorReg(int index) const {
+    return {};
+}
+
+void ARM_Dynarmic_32::SetVectorReg(int index, u128 value) {}
+
+u32 ARM_Dynarmic_32::GetPSTATE() const {
+    return jit->Cpsr();
+}
+
+void ARM_Dynarmic_32::SetPSTATE(u32 cpsr) {
+    jit->SetCpsr(cpsr);
+}
+
+u64 ARM_Dynarmic_32::GetTlsAddress() const {
+    return CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
+}
+
+void ARM_Dynarmic_32::SetTlsAddress(VAddr address) {
+    CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)] = static_cast<u32>(address);
+}
+
+u64 ARM_Dynarmic_32::GetTPIDR_EL0() const {
+    return cb->tpidr_el0;
+}
+
+void ARM_Dynarmic_32::SetTPIDR_EL0(u64 value) {
+    cb->tpidr_el0 = value;
+}
+
+void ARM_Dynarmic_32::SaveContext(ThreadContext32& ctx) {
+    Dynarmic::A32::Context context;
+    jit->SaveContext(context);
+    ctx.cpu_registers = context.Regs();
+    ctx.cpsr = context.Cpsr();
+}
+
+void ARM_Dynarmic_32::LoadContext(const ThreadContext32& ctx) {
+    Dynarmic::A32::Context context;
+    context.Regs() = ctx.cpu_registers;
+    context.SetCpsr(ctx.cpsr);
+    jit->LoadContext(context);
+}
+
+void ARM_Dynarmic_32::PrepareReschedule() {
+    jit->HaltExecution();
+}
+
+void ARM_Dynarmic_32::ClearInstructionCache() {
+    jit->ClearCache();
+}
+
+void ARM_Dynarmic_32::ClearExclusiveState() {}
+
+void ARM_Dynarmic_32::PageTableChanged(Common::PageTable& page_table,
+                                       std::size_t new_address_space_size_in_bits) {
+    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
+    auto iter = jit_cache.find(key);
+    if (iter != jit_cache.end()) {
+        jit = iter->second;
+        return;
+    }
+    jit = MakeJit(page_table, new_address_space_size_in_bits);
+    jit_cache.emplace(key, jit);
+}
+
+} // namespace Core
diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.h b/src/core/arm/dynarmic/arm_dynarmic_32.h
new file mode 100644
index 000000000..143e46e4d
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_32.h
@@ -0,0 +1,77 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <unordered_map>
+
+#include <dynarmic/A32/a32.h>
+#include <dynarmic/A64/a64.h>
+#include <dynarmic/A64/exclusive_monitor.h>
+#include "common/common_types.h"
+#include "common/hash.h"
+#include "core/arm/arm_interface.h"
+#include "core/arm/exclusive_monitor.h"
+
+namespace Memory {
+class Memory;
+}
+
+namespace Core {
+
+class DynarmicCallbacks32;
+class DynarmicExclusiveMonitor;
+class System;
+
+class ARM_Dynarmic_32 final : public ARM_Interface {
+public:
+    ARM_Dynarmic_32(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ~ARM_Dynarmic_32() override;
+
+    void SetPC(u64 pc) override;
+    u64 GetPC() const override;
+    u64 GetReg(int index) const override;
+    void SetReg(int index, u64 value) override;
+    u128 GetVectorReg(int index) const override;
+    void SetVectorReg(int index, u128 value) override;
+    u32 GetPSTATE() const override;
+    void SetPSTATE(u32 pstate) override;
+    void Run() override;
+    void Step() override;
+    VAddr GetTlsAddress() const override;
+    void SetTlsAddress(VAddr address) override;
+    void SetTPIDR_EL0(u64 value) override;
+    u64 GetTPIDR_EL0() const override;
+
+    void SaveContext(ThreadContext32& ctx) override;
+    void SaveContext(ThreadContext64& ctx) override {}
+    void LoadContext(const ThreadContext32& ctx) override;
+    void LoadContext(const ThreadContext64& ctx) override {}
+
+    void PrepareReschedule() override;
+    void ClearExclusiveState() override;
+
+    void ClearInstructionCache() override;
+    void PageTableChanged(Common::PageTable& new_page_table,
+                          std::size_t new_address_space_size_in_bits) override;
+
+private:
+    std::shared_ptr<Dynarmic::A32::Jit> MakeJit(Common::PageTable& page_table,
+                                                std::size_t address_space_bits) const;
+
+    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
+    using JitCacheType =
+        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A32::Jit>, Common::PairHash>;
+
+    friend class DynarmicCallbacks32;
+    std::unique_ptr<DynarmicCallbacks32> cb;
+    JitCacheType jit_cache;
+    std::shared_ptr<Dynarmic::A32::Jit> jit;
+    std::size_t core_index;
+    DynarmicExclusiveMonitor& exclusive_monitor;
+    std::array<u32, 84> CP15_regs{};
+};
+
+} // namespace Core
diff --git a/src/core/arm/dynarmic/arm_dynarmic.cpp b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
index 29eaf74e5..a53a58ba0 100644
--- a/src/core/arm/dynarmic/arm_dynarmic.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
@@ -8,7 +8,7 @@
 #include <dynarmic/A64/config.h>
 #include "common/logging/log.h"
 #include "common/microprofile.h"
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #include "core/core.h"
 #include "core/core_manager.h"
 #include "core/core_timing.h"
@@ -25,9 +25,9 @@ namespace Core {
 
 using Vector = Dynarmic::A64::Vector;
 
-class ARM_Dynarmic_Callbacks : public Dynarmic::A64::UserCallbacks {
+class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
-    explicit ARM_Dynarmic_Callbacks(ARM_Dynarmic& parent) : parent(parent) {}
+    explicit DynarmicCallbacks64(ARM_Dynarmic_64& parent) : parent(parent) {}
 
     u8 MemoryRead8(u64 vaddr) override {
         return parent.system.Memory().Read8(vaddr);
@@ -68,7 +68,7 @@ public:
         LOG_INFO(Core_ARM, "Unicorn fallback @ 0x{:X} for {} instructions (instr = {:08X})", pc,
                  num_instructions, MemoryReadCode(pc));
 
-        ARM_Interface::ThreadContext ctx;
+        ARM_Interface::ThreadContext64 ctx;
         parent.SaveContext(ctx);
         parent.inner_unicorn.LoadContext(ctx);
         parent.inner_unicorn.ExecuteInstructions(num_instructions);
@@ -90,7 +90,7 @@ public:
                 parent.jit->HaltExecution();
                 parent.SetPC(pc);
                 Kernel::Thread* const thread = parent.system.CurrentScheduler().GetCurrentThread();
-                parent.SaveContext(thread->GetContext());
+                parent.SaveContext(thread->GetContext64());
                 GDBStub::Break();
                 GDBStub::SendTrap(thread, 5);
                 return;
@@ -126,14 +126,14 @@ public:
         return Timing::CpuCyclesToClockCycles(parent.system.CoreTiming().GetTicks());
     }
 
-    ARM_Dynarmic& parent;
+    ARM_Dynarmic_64& parent;
     std::size_t num_interpreted_instructions = 0;
     u64 tpidrro_el0 = 0;
     u64 tpidr_el0 = 0;
 };
 
-std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit(Common::PageTable& page_table,
-                                                          std::size_t address_space_bits) const {
+std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable& page_table,
+                                                             std::size_t address_space_bits) const {
     Dynarmic::A64::UserConfig config;
 
     // Callbacks
@@ -159,79 +159,79 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit(Common::PageTable& pag
     // Unpredictable instructions
     config.define_unpredictable_behaviour = true;
 
-    return std::make_unique<Dynarmic::A64::Jit>(config);
+    return std::make_shared<Dynarmic::A64::Jit>(config);
 }
 
-MICROPROFILE_DEFINE(ARM_Jit_Dynarmic, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
+MICROPROFILE_DEFINE(ARM_Jit_Dynarmic_64, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
 
-void ARM_Dynarmic::Run() {
-    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic);
+void ARM_Dynarmic_64::Run() {
+    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic_64);
 
     jit->Run();
 }
 
-void ARM_Dynarmic::Step() {
+void ARM_Dynarmic_64::Step() {
     cb->InterpreterFallback(jit->GetPC(), 1);
 }
 
-ARM_Dynarmic::ARM_Dynarmic(System& system, ExclusiveMonitor& exclusive_monitor,
-                           std::size_t core_index)
+ARM_Dynarmic_64::ARM_Dynarmic_64(System& system, ExclusiveMonitor& exclusive_monitor,
+                                 std::size_t core_index)
     : ARM_Interface{system},
-      cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), inner_unicorn{system},
+      cb(std::make_unique<DynarmicCallbacks64>(*this)), inner_unicorn{system},
       core_index{core_index}, exclusive_monitor{
                                   dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
 
-ARM_Dynarmic::~ARM_Dynarmic() = default;
+ARM_Dynarmic_64::~ARM_Dynarmic_64() = default;
 
-void ARM_Dynarmic::SetPC(u64 pc) {
+void ARM_Dynarmic_64::SetPC(u64 pc) {
     jit->SetPC(pc);
 }
 
-u64 ARM_Dynarmic::GetPC() const {
+u64 ARM_Dynarmic_64::GetPC() const {
     return jit->GetPC();
 }
 
-u64 ARM_Dynarmic::GetReg(int index) const {
+u64 ARM_Dynarmic_64::GetReg(int index) const {
     return jit->GetRegister(index);
 }
 
-void ARM_Dynarmic::SetReg(int index, u64 value) {
+void ARM_Dynarmic_64::SetReg(int index, u64 value) {
     jit->SetRegister(index, value);
 }
 
-u128 ARM_Dynarmic::GetVectorReg(int index) const {
+u128 ARM_Dynarmic_64::GetVectorReg(int index) const {
     return jit->GetVector(index);
 }
 
-void ARM_Dynarmic::SetVectorReg(int index, u128 value) {
+void ARM_Dynarmic_64::SetVectorReg(int index, u128 value) {
     jit->SetVector(index, value);
 }
 
-u32 ARM_Dynarmic::GetPSTATE() const {
+u32 ARM_Dynarmic_64::GetPSTATE() const {
     return jit->GetPstate();
 }
 
-void ARM_Dynarmic::SetPSTATE(u32 pstate) {
+void ARM_Dynarmic_64::SetPSTATE(u32 pstate) {
     jit->SetPstate(pstate);
 }
 
-u64 ARM_Dynarmic::GetTlsAddress() const {
+u64 ARM_Dynarmic_64::GetTlsAddress() const {
     return cb->tpidrro_el0;
 }
 
-void ARM_Dynarmic::SetTlsAddress(VAddr address) {
+void ARM_Dynarmic_64::SetTlsAddress(VAddr address) {
     cb->tpidrro_el0 = address;
 }
 
-u64 ARM_Dynarmic::GetTPIDR_EL0() const {
+u64 ARM_Dynarmic_64::GetTPIDR_EL0() const {
     return cb->tpidr_el0;
 }
 
-void ARM_Dynarmic::SetTPIDR_EL0(u64 value) {
+void ARM_Dynarmic_64::SetTPIDR_EL0(u64 value) {
     cb->tpidr_el0 = value;
 }
 
-void ARM_Dynarmic::SaveContext(ThreadContext& ctx) {
+void ARM_Dynarmic_64::SaveContext(ThreadContext64& ctx) {
     ctx.cpu_registers = jit->GetRegisters();
     ctx.sp = jit->GetSP();
     ctx.pc = jit->GetPC();
@@ -242,7 +242,7 @@ void ARM_Dynarmic::SaveContext(ThreadContext& ctx) {
     ctx.tpidr = cb->tpidr_el0;
 }
 
-void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
+void ARM_Dynarmic_64::LoadContext(const ThreadContext64& ctx) {
     jit->SetRegisters(ctx.cpu_registers);
     jit->SetSP(ctx.sp);
     jit->SetPC(ctx.pc);
@@ -253,25 +253,32 @@ void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
     SetTPIDR_EL0(ctx.tpidr);
 }
 
-void ARM_Dynarmic::PrepareReschedule() {
+void ARM_Dynarmic_64::PrepareReschedule() {
     jit->HaltExecution();
 }
 
-void ARM_Dynarmic::ClearInstructionCache() {
+void ARM_Dynarmic_64::ClearInstructionCache() {
     jit->ClearCache();
 }
 
-void ARM_Dynarmic::ClearExclusiveState() {
+void ARM_Dynarmic_64::ClearExclusiveState() {
     jit->ClearExclusiveState();
 }
 
-void ARM_Dynarmic::PageTableChanged(Common::PageTable& page_table,
-                                    std::size_t new_address_space_size_in_bits) {
+void ARM_Dynarmic_64::PageTableChanged(Common::PageTable& page_table,
+                                       std::size_t new_address_space_size_in_bits) {
+    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
+    auto iter = jit_cache.find(key);
+    if (iter != jit_cache.end()) {
+        jit = iter->second;
+        return;
+    }
     jit = MakeJit(page_table, new_address_space_size_in_bits);
+    jit_cache.emplace(key, jit);
 }
 
-DynarmicExclusiveMonitor::DynarmicExclusiveMonitor(Memory::Memory& memory_, std::size_t core_count)
-    : monitor(core_count), memory{memory_} {}
+DynarmicExclusiveMonitor::DynarmicExclusiveMonitor(Memory::Memory& memory, std::size_t core_count)
+    : monitor(core_count), memory{memory} {}
 
 DynarmicExclusiveMonitor::~DynarmicExclusiveMonitor() = default;
 
diff --git a/src/core/arm/dynarmic/arm_dynarmic.h b/src/core/arm/dynarmic/arm_dynarmic_64.h
index 9cd475cfb..e71240a96 100644
--- a/src/core/arm/dynarmic/arm_dynarmic.h
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.h
@@ -5,9 +5,12 @@
 #pragma once
 
 #include <memory>
+#include <unordered_map>
+
 #include <dynarmic/A64/a64.h>
 #include <dynarmic/A64/exclusive_monitor.h>
 #include "common/common_types.h"
+#include "common/hash.h"
 #include "core/arm/arm_interface.h"
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
@@ -18,14 +21,14 @@ class Memory;
 
 namespace Core {
 
-class ARM_Dynarmic_Callbacks;
+class DynarmicCallbacks64;
 class DynarmicExclusiveMonitor;
 class System;
 
-class ARM_Dynarmic final : public ARM_Interface {
+class ARM_Dynarmic_64 final : public ARM_Interface {
 public:
-    ARM_Dynarmic(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
-    ~ARM_Dynarmic() override;
+    ARM_Dynarmic_64(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ~ARM_Dynarmic_64() override;
 
     void SetPC(u64 pc) override;
     u64 GetPC() const override;
@@ -42,8 +45,10 @@ public:
     void SetTPIDR_EL0(u64 value) override;
     u64 GetTPIDR_EL0() const override;
 
-    void SaveContext(ThreadContext& ctx) override;
-    void LoadContext(const ThreadContext& ctx) override;
+    void SaveContext(ThreadContext32& ctx) override {}
+    void SaveContext(ThreadContext64& ctx) override;
+    void LoadContext(const ThreadContext32& ctx) override {}
+    void LoadContext(const ThreadContext64& ctx) override;
 
     void PrepareReschedule() override;
     void ClearExclusiveState() override;
@@ -53,12 +58,17 @@ public:
                           std::size_t new_address_space_size_in_bits) override;
 
 private:
-    std::unique_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable& page_table,
+    std::shared_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable& page_table,
                                                 std::size_t address_space_bits) const;
 
-    friend class ARM_Dynarmic_Callbacks;
-    std::unique_ptr<ARM_Dynarmic_Callbacks> cb;
-    std::unique_ptr<Dynarmic::A64::Jit> jit;
+    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
+    using JitCacheType =
+        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A64::Jit>, Common::PairHash>;
+
+    friend class DynarmicCallbacks64;
+    std::unique_ptr<DynarmicCallbacks64> cb;
+    JitCacheType jit_cache;
+    std::shared_ptr<Dynarmic::A64::Jit> jit;
     ARM_Unicorn inner_unicorn;
 
     std::size_t core_index;
@@ -67,7 +77,7 @@ private:
 
 class DynarmicExclusiveMonitor final : public ExclusiveMonitor {
 public:
-    explicit DynarmicExclusiveMonitor(Memory::Memory& memory_, std::size_t core_count);
+    explicit DynarmicExclusiveMonitor(Memory::Memory& memory, std::size_t core_count);
     ~DynarmicExclusiveMonitor() override;
 
     void SetExclusive(std::size_t core_index, VAddr addr) override;
@@ -80,7 +90,7 @@ public:
     bool ExclusiveWrite128(std::size_t core_index, VAddr vaddr, u128 value) override;
 
 private:
-    friend class ARM_Dynarmic;
+    friend class ARM_Dynarmic_64;
     Dynarmic::A64::ExclusiveMonitor monitor;
     Memory::Memory& memory;
 };
diff --git a/src/core/arm/dynarmic/arm_dynarmic_cp15.cpp b/src/core/arm/dynarmic/arm_dynarmic_cp15.cpp
new file mode 100644
index 000000000..3fdcdebde
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_cp15.cpp
@@ -0,0 +1,80 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/arm/dynarmic/arm_dynarmic_cp15.h"
+
+using Callback = Dynarmic::A32::Coprocessor::Callback;
+using CallbackOrAccessOneWord = Dynarmic::A32::Coprocessor::CallbackOrAccessOneWord;
+using CallbackOrAccessTwoWords = Dynarmic::A32::Coprocessor::CallbackOrAccessTwoWords;
+
+std::optional<Callback> DynarmicCP15::CompileInternalOperation(bool two, unsigned opc1,
+                                                               CoprocReg CRd, CoprocReg CRn,
+                                                               CoprocReg CRm, unsigned opc2) {
+    return {};
+}
+
+CallbackOrAccessOneWord DynarmicCP15::CompileSendOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                                         CoprocReg CRm, unsigned opc2) {
+    // TODO(merry): Privileged CP15 registers
+
+    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C5 && opc2 == 4) {
+        // This is a dummy write, we ignore the value written here.
+        return &CP15[static_cast<std::size_t>(CP15Register::CP15_FLUSH_PREFETCH_BUFFER)];
+    }
+
+    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C10) {
+        switch (opc2) {
+        case 4:
+            // This is a dummy write, we ignore the value written here.
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_SYNC_BARRIER)];
+        case 5:
+            // This is a dummy write, we ignore the value written here.
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_MEMORY_BARRIER)];
+        default:
+            return {};
+        }
+    }
+
+    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0 && opc2 == 2) {
+        return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+    }
+
+    return {};
+}
+
+CallbackOrAccessTwoWords DynarmicCP15::CompileSendTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    return {};
+}
+
+CallbackOrAccessOneWord DynarmicCP15::CompileGetOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                                        CoprocReg CRm, unsigned opc2) {
+    // TODO(merry): Privileged CP15 registers
+
+    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0) {
+        switch (opc2) {
+        case 2:
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+        case 3:
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
+        default:
+            return {};
+        }
+    }
+
+    return {};
+}
+
+CallbackOrAccessTwoWords DynarmicCP15::CompileGetTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    return {};
+}
+
+std::optional<Callback> DynarmicCP15::CompileLoadWords(bool two, bool long_transfer, CoprocReg CRd,
+                                                       std::optional<u8> option) {
+    return {};
+}
+
+std::optional<Callback> DynarmicCP15::CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
+                                                        std::optional<u8> option) {
+    return {};
+}
diff --git a/src/core/arm/dynarmic/arm_dynarmic_cp15.h b/src/core/arm/dynarmic/arm_dynarmic_cp15.h
new file mode 100644
index 000000000..07bcde5f9
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_cp15.h
@@ -0,0 +1,152 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <optional>
+
+#include <dynarmic/A32/coprocessor.h>
+#include "common/common_types.h"
+
+enum class CP15Register {
+    // c0 - Information registers
+    CP15_MAIN_ID,
+    CP15_CACHE_TYPE,
+    CP15_TCM_STATUS,
+    CP15_TLB_TYPE,
+    CP15_CPU_ID,
+    CP15_PROCESSOR_FEATURE_0,
+    CP15_PROCESSOR_FEATURE_1,
+    CP15_DEBUG_FEATURE_0,
+    CP15_AUXILIARY_FEATURE_0,
+    CP15_MEMORY_MODEL_FEATURE_0,
+    CP15_MEMORY_MODEL_FEATURE_1,
+    CP15_MEMORY_MODEL_FEATURE_2,
+    CP15_MEMORY_MODEL_FEATURE_3,
+    CP15_ISA_FEATURE_0,
+    CP15_ISA_FEATURE_1,
+    CP15_ISA_FEATURE_2,
+    CP15_ISA_FEATURE_3,
+    CP15_ISA_FEATURE_4,
+
+    // c1 - Control registers
+    CP15_CONTROL,
+    CP15_AUXILIARY_CONTROL,
+    CP15_COPROCESSOR_ACCESS_CONTROL,
+
+    // c2 - Translation table registers
+    CP15_TRANSLATION_BASE_TABLE_0,
+    CP15_TRANSLATION_BASE_TABLE_1,
+    CP15_TRANSLATION_BASE_CONTROL,
+    CP15_DOMAIN_ACCESS_CONTROL,
+    CP15_RESERVED,
+
+    // c5 - Fault status registers
+    CP15_FAULT_STATUS,
+    CP15_INSTR_FAULT_STATUS,
+    CP15_COMBINED_DATA_FSR = CP15_FAULT_STATUS,
+    CP15_INST_FSR,
+
+    // c6 - Fault Address registers
+    CP15_FAULT_ADDRESS,
+    CP15_COMBINED_DATA_FAR = CP15_FAULT_ADDRESS,
+    CP15_WFAR,
+    CP15_IFAR,
+
+    // c7 - Cache operation registers
+    CP15_WAIT_FOR_INTERRUPT,
+    CP15_PHYS_ADDRESS,
+    CP15_INVALIDATE_INSTR_CACHE,
+    CP15_INVALIDATE_INSTR_CACHE_USING_MVA,
+    CP15_INVALIDATE_INSTR_CACHE_USING_INDEX,
+    CP15_FLUSH_PREFETCH_BUFFER,
+    CP15_FLUSH_BRANCH_TARGET_CACHE,
+    CP15_FLUSH_BRANCH_TARGET_CACHE_ENTRY,
+    CP15_INVALIDATE_DATA_CACHE,
+    CP15_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
+    CP15_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
+    CP15_INVALIDATE_DATA_AND_INSTR_CACHE,
+    CP15_CLEAN_DATA_CACHE,
+    CP15_CLEAN_DATA_CACHE_LINE_USING_MVA,
+    CP15_CLEAN_DATA_CACHE_LINE_USING_INDEX,
+    CP15_DATA_SYNC_BARRIER,
+    CP15_DATA_MEMORY_BARRIER,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
+
+    // c8 - TLB operations
+    CP15_INVALIDATE_ITLB,
+    CP15_INVALIDATE_ITLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_ITLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_ITLB_ENTRY_ON_MVA,
+    CP15_INVALIDATE_DTLB,
+    CP15_INVALIDATE_DTLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_DTLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_DTLB_ENTRY_ON_MVA,
+    CP15_INVALIDATE_UTLB,
+    CP15_INVALIDATE_UTLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_UTLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_UTLB_ENTRY_ON_MVA,
+
+    // c9 - Data cache lockdown register
+    CP15_DATA_CACHE_LOCKDOWN,
+
+    // c10 - TLB/Memory map registers
+    CP15_TLB_LOCKDOWN,
+    CP15_PRIMARY_REGION_REMAP,
+    CP15_NORMAL_REGION_REMAP,
+
+    // c13 - Thread related registers
+    CP15_PID,
+    CP15_CONTEXT_ID,
+    CP15_THREAD_UPRW, // Thread ID register - User/Privileged Read/Write
+    CP15_THREAD_URO,  // Thread ID register - User Read Only (Privileged R/W)
+    CP15_THREAD_PRW,  // Thread ID register - Privileged R/W only.
+
+    // c15 - Performance and TLB lockdown registers
+    CP15_PERFORMANCE_MONITOR_CONTROL,
+    CP15_CYCLE_COUNTER,
+    CP15_COUNT_0,
+    CP15_COUNT_1,
+    CP15_READ_MAIN_TLB_LOCKDOWN_ENTRY,
+    CP15_WRITE_MAIN_TLB_LOCKDOWN_ENTRY,
+    CP15_MAIN_TLB_LOCKDOWN_VIRT_ADDRESS,
+    CP15_MAIN_TLB_LOCKDOWN_PHYS_ADDRESS,
+    CP15_MAIN_TLB_LOCKDOWN_ATTRIBUTE,
+    CP15_TLB_DEBUG_CONTROL,
+
+    // Skyeye defined
+    CP15_TLB_FAULT_ADDR,
+    CP15_TLB_FAULT_STATUS,
+
+    // Not an actual register.
+    // All registers should be defined above this.
+    CP15_REGISTER_COUNT,
+};
+
+class DynarmicCP15 final : public Dynarmic::A32::Coprocessor {
+public:
+    using CoprocReg = Dynarmic::A32::CoprocReg;
+
+    explicit DynarmicCP15(u32* cp15) : CP15(cp15){};
+
+    std::optional<Callback> CompileInternalOperation(bool two, unsigned opc1, CoprocReg CRd,
+                                                     CoprocReg CRn, CoprocReg CRm,
+                                                     unsigned opc2) override;
+    CallbackOrAccessOneWord CompileSendOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                               CoprocReg CRm, unsigned opc2) override;
+    CallbackOrAccessTwoWords CompileSendTwoWords(bool two, unsigned opc, CoprocReg CRm) override;
+    CallbackOrAccessOneWord CompileGetOneWord(bool two, unsigned opc1, CoprocReg CRn, CoprocReg CRm,
+                                              unsigned opc2) override;
+    CallbackOrAccessTwoWords CompileGetTwoWords(bool two, unsigned opc, CoprocReg CRm) override;
+    std::optional<Callback> CompileLoadWords(bool two, bool long_transfer, CoprocReg CRd,
+                                             std::optional<u8> option) override;
+    std::optional<Callback> CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
+                                              std::optional<u8> option) override;
+
+private:
+    u32* CP15{};
+};
diff --git a/src/core/arm/exclusive_monitor.cpp b/src/core/arm/exclusive_monitor.cpp
index 94570e520..b32401e0b 100644
--- a/src/core/arm/exclusive_monitor.cpp
+++ b/src/core/arm/exclusive_monitor.cpp
@@ -3,7 +3,7 @@
 // Refer to the license.txt file included.
 
 #ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/memory.h"
diff --git a/src/core/arm/unicorn/arm_unicorn.cpp b/src/core/arm/unicorn/arm_unicorn.cpp
index f99ad5802..8a9800a96 100644
--- a/src/core/arm/unicorn/arm_unicorn.cpp
+++ b/src/core/arm/unicorn/arm_unicorn.cpp
@@ -53,7 +53,7 @@ static bool UnmappedMemoryHook(uc_engine* uc, uc_mem_type type, u64 addr, int si
                                void* user_data) {
     auto* const system = static_cast<System*>(user_data);
 
-    ARM_Interface::ThreadContext ctx{};
+    ARM_Interface::ThreadContext64 ctx{};
     system->CurrentArmInterface().SaveContext(ctx);
     ASSERT_MSG(false, "Attempted to read from unmapped memory: 0x{:X}, pc=0x{:X}, lr=0x{:X}", addr,
                ctx.pc, ctx.cpu_registers[30]);
@@ -179,7 +179,7 @@ void ARM_Unicorn::ExecuteInstructions(std::size_t num_instructions) {
         }
 
         Kernel::Thread* const thread = system.CurrentScheduler().GetCurrentThread();
-        SaveContext(thread->GetContext());
+        SaveContext(thread->GetContext64());
         if (last_bkpt_hit || GDBStub::IsMemoryBreak() || GDBStub::GetCpuStepFlag()) {
             last_bkpt_hit = false;
             GDBStub::Break();
@@ -188,7 +188,7 @@ void ARM_Unicorn::ExecuteInstructions(std::size_t num_instructions) {
     }
 }
 
-void ARM_Unicorn::SaveContext(ThreadContext& ctx) {
+void ARM_Unicorn::SaveContext(ThreadContext64& ctx) {
     int uregs[32];
     void* tregs[32];
 
@@ -215,7 +215,7 @@ void ARM_Unicorn::SaveContext(ThreadContext& ctx) {
     CHECKED(uc_reg_read_batch(uc, uregs, tregs, 32));
 }
 
-void ARM_Unicorn::LoadContext(const ThreadContext& ctx) {
+void ARM_Unicorn::LoadContext(const ThreadContext64& ctx) {
     int uregs[32];
     void* tregs[32];
 
diff --git a/src/core/arm/unicorn/arm_unicorn.h b/src/core/arm/unicorn/arm_unicorn.h
index 3c5b155f9..f30d13cb6 100644
--- a/src/core/arm/unicorn/arm_unicorn.h
+++ b/src/core/arm/unicorn/arm_unicorn.h
@@ -30,8 +30,6 @@ public:
     void SetTlsAddress(VAddr address) override;
     void SetTPIDR_EL0(u64 value) override;
     u64 GetTPIDR_EL0() const override;
-    void SaveContext(ThreadContext& ctx) override;
-    void LoadContext(const ThreadContext& ctx) override;
     void PrepareReschedule() override;
     void ClearExclusiveState() override;
     void ExecuteInstructions(std::size_t num_instructions);
@@ -41,6 +39,11 @@ public:
     void PageTableChanged(Common::PageTable&, std::size_t) override {}
     void RecordBreak(GDBStub::BreakpointAddress bkpt);
 
+    void SaveContext(ThreadContext32& ctx) override {}
+    void SaveContext(ThreadContext64& ctx) override;
+    void LoadContext(const ThreadContext32& ctx) override {}
+    void LoadContext(const ThreadContext64& ctx) override;
+
 private:
     static void InterruptHook(uc_engine* uc, u32 int_no, void* user_data);
 
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 86e314c94..d1bc9340d 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -24,6 +24,7 @@
 #include "core/file_sys/sdmc_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hardware_interrupt_manager.h"
 #include "core/hle/kernel/client_port.h"
@@ -165,7 +166,7 @@ struct System::Impl {
         service_manager = std::make_shared<Service::SM::ServiceManager>();
 
         Service::Init(service_manager, system);
-        GDBStub::Init();
+        GDBStub::DeferStart();
 
         renderer = VideoCore::CreateRenderer(emu_window, system);
         if (!renderer->Init()) {
@@ -173,6 +174,7 @@ struct System::Impl {
         }
         interrupt_manager = std::make_unique<Core::Hardware::InterruptManager>(system);
         gpu_core = VideoCore::CreateGPU(system);
+        renderer->Rasterizer().SetupDirtyFlags();
 
         is_powered_on = true;
         exit_lock = false;
@@ -184,6 +186,8 @@ struct System::Impl {
 
     ResultStatus Load(System& system, Frontend::EmuWindow& emu_window,
                       const std::string& filepath) {
+        Core::Frontend::ScopeAcquireContext acquire_context{emu_window};
+
         app_loader = Loader::GetLoader(GetGameFileFromPath(virtual_filesystem, filepath));
         if (!app_loader) {
             LOG_CRITICAL(Core, "Failed to obtain loader for {}!", filepath);
diff --git a/src/core/core_manager.cpp b/src/core/core_manager.cpp
index 8eacf92dd..b6b797c80 100644
--- a/src/core/core_manager.cpp
+++ b/src/core/core_manager.cpp
@@ -6,9 +6,6 @@
 #include <mutex>
 
 #include "common/logging/log.h"
-#ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
-#endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
 #include "core/core.h"
diff --git a/src/core/frontend/emu_window.h b/src/core/frontend/emu_window.h
index 3376eedc5..5eb87fb63 100644
--- a/src/core/frontend/emu_window.h
+++ b/src/core/frontend/emu_window.h
@@ -26,9 +26,6 @@ public:
 
     /// Releases (dunno if this is the "right" word) the context from the caller thread
     virtual void DoneCurrent() = 0;
-
-    /// Swap buffers to display the next frame
-    virtual void SwapBuffers() = 0;
 };
 
 /**
diff --git a/src/core/frontend/framebuffer_layout.cpp b/src/core/frontend/framebuffer_layout.cpp
index 2dc795d56..68a0e0906 100644
--- a/src/core/frontend/framebuffer_layout.cpp
+++ b/src/core/frontend/framebuffer_layout.cpp
@@ -48,8 +48,8 @@ FramebufferLayout FrameLayoutFromResolutionScale(u32 res_scale) {
     u32 width, height;
 
     if (Settings::values.use_docked_mode) {
-        width = ScreenDocked::WidthDocked * res_scale;
-        height = ScreenDocked::HeightDocked * res_scale;
+        width = ScreenDocked::Width * res_scale;
+        height = ScreenDocked::Height * res_scale;
     } else {
         width = ScreenUndocked::Width * res_scale;
         height = ScreenUndocked::Height * res_scale;
diff --git a/src/core/frontend/framebuffer_layout.h b/src/core/frontend/framebuffer_layout.h
index 1d39c1faf..15ecfb13d 100644
--- a/src/core/frontend/framebuffer_layout.h
+++ b/src/core/frontend/framebuffer_layout.h
@@ -8,15 +8,15 @@
 
 namespace Layout {
 
-enum ScreenUndocked : u32 {
-    Width = 1280,
-    Height = 720,
-};
+namespace ScreenUndocked {
+constexpr u32 Width = 1280;
+constexpr u32 Height = 720;
+} // namespace ScreenUndocked
 
-enum ScreenDocked : u32 {
-    WidthDocked = 1920,
-    HeightDocked = 1080,
-};
+namespace ScreenDocked {
+constexpr u32 Width = 1920;
+constexpr u32 Height = 1080;
+} // namespace ScreenDocked
 
 enum class AspectRatio {
     Default,
@@ -29,6 +29,7 @@ enum class AspectRatio {
 struct FramebufferLayout {
     u32 width{ScreenUndocked::Width};
     u32 height{ScreenUndocked::Height};
+    bool is_srgb{};
 
     Common::Rectangle<u32> screen;
 
diff --git a/src/core/frontend/scope_acquire_context.cpp b/src/core/frontend/scope_acquire_context.cpp
new file mode 100644
index 000000000..878c3157c
--- /dev/null
+++ b/src/core/frontend/scope_acquire_context.cpp
@@ -0,0 +1,18 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/frontend/emu_window.h"
+#include "core/frontend/scope_acquire_context.h"
+
+namespace Core::Frontend {
+
+ScopeAcquireContext::ScopeAcquireContext(Core::Frontend::GraphicsContext& context)
+    : context{context} {
+    context.MakeCurrent();
+}
+ScopeAcquireContext::~ScopeAcquireContext() {
+    context.DoneCurrent();
+}
+
+} // namespace Core::Frontend
diff --git a/src/core/frontend/scope_acquire_window_context.h b/src/core/frontend/scope_acquire_context.h
index 2d9f6e825..7a65c0623 100644
--- a/src/core/frontend/scope_acquire_window_context.h
+++ b/src/core/frontend/scope_acquire_context.h
@@ -8,16 +8,16 @@
 
 namespace Core::Frontend {
 
-class EmuWindow;
+class GraphicsContext;
 
 /// Helper class to acquire/release window context within a given scope
-class ScopeAcquireWindowContext : NonCopyable {
+class ScopeAcquireContext : NonCopyable {
 public:
-    explicit ScopeAcquireWindowContext(Core::Frontend::EmuWindow& window);
-    ~ScopeAcquireWindowContext();
+    explicit ScopeAcquireContext(Core::Frontend::GraphicsContext& context);
+    ~ScopeAcquireContext();
 
 private:
-    Core::Frontend::EmuWindow& emu_window;
+    Core::Frontend::GraphicsContext& context;
 };
 
 } // namespace Core::Frontend
diff --git a/src/core/frontend/scope_acquire_window_context.cpp b/src/core/frontend/scope_acquire_window_context.cpp
deleted file mode 100644
index 3663dad17..000000000
--- a/src/core/frontend/scope_acquire_window_context.cpp
+++ /dev/null
@@ -1,18 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_window_context.h"
-
-namespace Core::Frontend {
-
-ScopeAcquireWindowContext::ScopeAcquireWindowContext(Core::Frontend::EmuWindow& emu_window_)
-    : emu_window{emu_window_} {
-    emu_window.MakeCurrent();
-}
-ScopeAcquireWindowContext::~ScopeAcquireWindowContext() {
-    emu_window.DoneCurrent();
-}
-
-} // namespace Core::Frontend
diff --git a/src/core/gdbstub/gdbstub.cpp b/src/core/gdbstub/gdbstub.cpp
index 67e95999d..6d15aeed9 100644
--- a/src/core/gdbstub/gdbstub.cpp
+++ b/src/core/gdbstub/gdbstub.cpp
@@ -141,6 +141,7 @@ constexpr char target_xml[] =
 )";
 
 int gdbserver_socket = -1;
+bool defer_start = false;
 
 u8 command_buffer[GDB_BUFFER_SIZE];
 u32 command_length;
@@ -217,7 +218,7 @@ static u64 RegRead(std::size_t id, Kernel::Thread* thread = nullptr) {
         return 0;
     }
 
-    const auto& thread_context = thread->GetContext();
+    const auto& thread_context = thread->GetContext64();
 
     if (id < SP_REGISTER) {
         return thread_context.cpu_registers[id];
@@ -239,7 +240,7 @@ static void RegWrite(std::size_t id, u64 val, Kernel::Thread* thread = nullptr)
         return;
     }
 
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
 
     if (id < SP_REGISTER) {
         thread_context.cpu_registers[id] = val;
@@ -259,7 +260,7 @@ static u128 FpuRead(std::size_t id, Kernel::Thread* thread = nullptr) {
         return u128{0};
     }
 
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
 
     if (id >= UC_ARM64_REG_Q0 && id < FPCR_REGISTER) {
         return thread_context.vector_registers[id - UC_ARM64_REG_Q0];
@@ -275,7 +276,7 @@ static void FpuWrite(std::size_t id, u128 val, Kernel::Thread* thread = nullptr)
         return;
     }
 
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
 
     if (id >= UC_ARM64_REG_Q0 && id < FPCR_REGISTER) {
         thread_context.vector_registers[id - UC_ARM64_REG_Q0] = val;
@@ -916,7 +917,7 @@ static void WriteRegister() {
     // Update ARM context, skipping scheduler - no running threads at this point
     Core::System::GetInstance()
         .ArmInterface(current_core)
-        .LoadContext(current_thread->GetContext());
+        .LoadContext(current_thread->GetContext64());
 
     SendReply("OK");
 }
@@ -947,7 +948,7 @@ static void WriteRegisters() {
     // Update ARM context, skipping scheduler - no running threads at this point
     Core::System::GetInstance()
         .ArmInterface(current_core)
-        .LoadContext(current_thread->GetContext());
+        .LoadContext(current_thread->GetContext64());
 
     SendReply("OK");
 }
@@ -1019,7 +1020,7 @@ static void Step() {
         // Update ARM context, skipping scheduler - no running threads at this point
         Core::System::GetInstance()
             .ArmInterface(current_core)
-            .LoadContext(current_thread->GetContext());
+            .LoadContext(current_thread->GetContext64());
     }
     step_loop = true;
     halt_loop = true;
@@ -1166,6 +1167,9 @@ static void RemoveBreakpoint() {
 
 void HandlePacket() {
     if (!IsConnected()) {
+        if (defer_start) {
+            ToggleServer(true);
+        }
         return;
     }
 
@@ -1256,6 +1260,10 @@ void ToggleServer(bool status) {
     }
 }
 
+void DeferStart() {
+    defer_start = true;
+}
+
 static void Init(u16 port) {
     if (!server_enabled) {
         // Set the halt loop to false in case the user enabled the gdbstub mid-execution.
@@ -1341,6 +1349,7 @@ void Shutdown() {
     if (!server_enabled) {
         return;
     }
+    defer_start = false;
 
     LOG_INFO(Debug_GDBStub, "Stopping GDB ...");
     if (gdbserver_socket != -1) {
diff --git a/src/core/gdbstub/gdbstub.h b/src/core/gdbstub/gdbstub.h
index 5a36524b2..8fe3c320b 100644
--- a/src/core/gdbstub/gdbstub.h
+++ b/src/core/gdbstub/gdbstub.h
@@ -43,6 +43,13 @@ void ToggleServer(bool status);
 /// Start the gdbstub server.
 void Init();
 
+/**
+ * Defer initialization of the gdbstub to the first packet processing functions.
+ * This avoids a case where the gdbstub thread is frozen after initialization
+ * and fails to respond in time to packets.
+ */
+void DeferStart();
+
 /// Stop gdbstub server.
 void Shutdown();
 
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 9232f4d7e..e47f1deed 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -186,6 +186,10 @@ struct KernelCore::Impl {
             return;
         }
 
+        for (auto& core : cores) {
+            core.SetIs64Bit(process->Is64BitProcess());
+        }
+
         system.Memory().SetCurrentPageTable(*process);
     }
 
diff --git a/src/core/hle/kernel/physical_core.cpp b/src/core/hle/kernel/physical_core.cpp
index 9303dd273..aa2787467 100644
--- a/src/core/hle/kernel/physical_core.cpp
+++ b/src/core/hle/kernel/physical_core.cpp
@@ -5,7 +5,8 @@
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
 #ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
@@ -20,13 +21,17 @@ PhysicalCore::PhysicalCore(Core::System& system, std::size_t id,
                            Core::ExclusiveMonitor& exclusive_monitor)
     : core_index{id} {
 #ifdef ARCHITECTURE_x86_64
-    arm_interface = std::make_unique<Core::ARM_Dynarmic>(system, exclusive_monitor, core_index);
+    arm_interface_32 =
+        std::make_unique<Core::ARM_Dynarmic_32>(system, exclusive_monitor, core_index);
+    arm_interface_64 =
+        std::make_unique<Core::ARM_Dynarmic_64>(system, exclusive_monitor, core_index);
+
 #else
     arm_interface = std::make_shared<Core::ARM_Unicorn>(system);
     LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
 
-    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface, core_index);
+    scheduler = std::make_unique<Kernel::Scheduler>(system, core_index);
 }
 
 PhysicalCore::~PhysicalCore() = default;
@@ -48,4 +53,12 @@ void PhysicalCore::Shutdown() {
     scheduler->Shutdown();
 }
 
+void PhysicalCore::SetIs64Bit(bool is_64_bit) {
+    if (is_64_bit) {
+        arm_interface = arm_interface_64.get();
+    } else {
+        arm_interface = arm_interface_32.get();
+    }
+}
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/physical_core.h b/src/core/hle/kernel/physical_core.h
index 4c32c0f1b..3269166be 100644
--- a/src/core/hle/kernel/physical_core.h
+++ b/src/core/hle/kernel/physical_core.h
@@ -68,10 +68,14 @@ public:
         return *scheduler;
     }
 
+    void SetIs64Bit(bool is_64_bit);
+
 private:
     std::size_t core_index;
-    std::unique_ptr<Core::ARM_Interface> arm_interface;
+    std::unique_ptr<Core::ARM_Interface> arm_interface_32;
+    std::unique_ptr<Core::ARM_Interface> arm_interface_64;
     std::unique_ptr<Kernel::Scheduler> scheduler;
+    Core::ARM_Interface* arm_interface{};
 };
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index 2fcb7326c..edc414d69 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -42,7 +42,8 @@ void SetupMainThread(Process& owner_process, KernelCore& kernel, u32 priority) {
 
     // Register 1 must be a handle to the main thread
     const Handle thread_handle = owner_process.GetHandleTable().Create(thread).Unwrap();
-    thread->GetContext().cpu_registers[1] = thread_handle;
+    thread->GetContext32().cpu_registers[1] = thread_handle;
+    thread->GetContext64().cpu_registers[1] = thread_handle;
 
     // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires
     thread->ResumeFromWait();
diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp
index c65f82fb7..1140c72a3 100644
--- a/src/core/hle/kernel/scheduler.cpp
+++ b/src/core/hle/kernel/scheduler.cpp
@@ -383,8 +383,8 @@ void GlobalScheduler::Unlock() {
     // TODO(Blinkhawk): Setup the interrupts and change context on current core.
 }
 
-Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, std::size_t core_id)
-    : system(system), cpu_core(cpu_core), core_id(core_id) {}
+Scheduler::Scheduler(Core::System& system, std::size_t core_id)
+    : system{system}, core_id{core_id} {}
 
 Scheduler::~Scheduler() = default;
 
@@ -422,9 +422,10 @@ void Scheduler::UnloadThread() {
 
     // Save context for previous thread
     if (previous_thread) {
-        cpu_core.SaveContext(previous_thread->GetContext());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext32());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext64());
         // Save the TPIDR_EL0 system register in case it was modified.
-        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+        previous_thread->SetTPIDR_EL0(system.ArmInterface(core_id).GetTPIDR_EL0());
 
         if (previous_thread->GetStatus() == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
@@ -451,9 +452,10 @@ void Scheduler::SwitchContext() {
 
     // Save context for previous thread
     if (previous_thread) {
-        cpu_core.SaveContext(previous_thread->GetContext());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext32());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext64());
         // Save the TPIDR_EL0 system register in case it was modified.
-        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+        previous_thread->SetTPIDR_EL0(system.ArmInterface(core_id).GetTPIDR_EL0());
 
         if (previous_thread->GetStatus() == ThreadStatus::Running) {
             // This is only the case when a reschedule is triggered without the current thread
@@ -481,9 +483,10 @@ void Scheduler::SwitchContext() {
             system.Kernel().MakeCurrentProcess(thread_owner_process);
         }
 
-        cpu_core.LoadContext(new_thread->GetContext());
-        cpu_core.SetTlsAddress(new_thread->GetTLSAddress());
-        cpu_core.SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
+        system.ArmInterface(core_id).LoadContext(new_thread->GetContext32());
+        system.ArmInterface(core_id).LoadContext(new_thread->GetContext64());
+        system.ArmInterface(core_id).SetTlsAddress(new_thread->GetTLSAddress());
+        system.ArmInterface(core_id).SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
     } else {
         current_thread = nullptr;
         // Note: We do not reset the current process and current page table when idling because
diff --git a/src/core/hle/kernel/scheduler.h b/src/core/hle/kernel/scheduler.h
index 1c93a838c..07df33f9c 100644
--- a/src/core/hle/kernel/scheduler.h
+++ b/src/core/hle/kernel/scheduler.h
@@ -181,7 +181,7 @@ private:
 
 class Scheduler final {
 public:
-    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, std::size_t core_id);
+    explicit Scheduler(Core::System& system, std::size_t core_id);
     ~Scheduler();
 
     /// Returns whether there are any threads that are ready to run.
@@ -235,7 +235,6 @@ private:
     std::shared_ptr<Thread> selected_thread = nullptr;
 
     Core::System& system;
-    Core::ARM_Interface& cpu_core;
     u64 last_context_switch_time = 0;
     u64 idle_selection_count = 0;
     const std::size_t core_id;
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index fd91779a3..4ffc113c2 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -187,6 +187,13 @@ static ResultCode SetHeapSize(Core::System& system, VAddr* heap_addr, u64 heap_s
     return RESULT_SUCCESS;
 }
 
+static ResultCode SetHeapSize32(Core::System& system, u32* heap_addr, u32 heap_size) {
+    VAddr temp_heap_addr{};
+    const ResultCode result{SetHeapSize(system, &temp_heap_addr, heap_size)};
+    *heap_addr = static_cast<u32>(temp_heap_addr);
+    return result;
+}
+
 static ResultCode SetMemoryPermission(Core::System& system, VAddr addr, u64 size, u32 prot) {
     LOG_TRACE(Kernel_SVC, "called, addr=0x{:X}, size=0x{:X}, prot=0x{:X}", addr, size, prot);
 
@@ -371,6 +378,12 @@ static ResultCode ConnectToNamedPort(Core::System& system, Handle* out_handle,
     return RESULT_SUCCESS;
 }
 
+static ResultCode ConnectToNamedPort32(Core::System& system, Handle* out_handle,
+                                       u32 port_name_address) {
+
+    return ConnectToNamedPort(system, out_handle, port_name_address);
+}
+
 /// Makes a blocking IPC call to an OS service.
 static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
     const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
@@ -390,6 +403,10 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
     return session->SendSyncRequest(SharedFrom(thread), system.Memory());
 }
 
+static ResultCode SendSyncRequest32(Core::System& system, Handle handle) {
+    return SendSyncRequest(system, handle);
+}
+
 /// Get the ID for the specified thread.
 static ResultCode GetThreadId(Core::System& system, u64* thread_id, Handle thread_handle) {
     LOG_TRACE(Kernel_SVC, "called thread=0x{:08X}", thread_handle);
@@ -405,6 +422,17 @@ static ResultCode GetThreadId(Core::System& system, u64* thread_id, Handle threa
     return RESULT_SUCCESS;
 }
 
+static ResultCode GetThreadId32(Core::System& system, u32* thread_id_low, u32* thread_id_high,
+                                Handle thread_handle) {
+    u64 thread_id{};
+    const ResultCode result{GetThreadId(system, &thread_id, thread_handle)};
+
+    *thread_id_low = static_cast<u32>(thread_id >> 32);
+    *thread_id_high = static_cast<u32>(thread_id & std::numeric_limits<u32>::max());
+
+    return result;
+}
+
 /// Gets the ID of the specified process or a specified thread's owning process.
 static ResultCode GetProcessId(Core::System& system, u64* process_id, Handle handle) {
     LOG_DEBUG(Kernel_SVC, "called handle=0x{:08X}", handle);
@@ -479,6 +507,12 @@ static ResultCode WaitSynchronization(Core::System& system, Handle* index, VAddr
     return result;
 }
 
+static ResultCode WaitSynchronization32(Core::System& system, u32 timeout_low, u32 handles_address,
+                                        s32 handle_count, u32 timeout_high, Handle* index) {
+    const s64 nano_seconds{(static_cast<s64>(timeout_high) << 32) | static_cast<s64>(timeout_low)};
+    return WaitSynchronization(system, index, handles_address, handle_count, nano_seconds);
+}
+
 /// Resumes a thread waiting on WaitSynchronization
 static ResultCode CancelSynchronization(Core::System& system, Handle thread_handle) {
     LOG_TRACE(Kernel_SVC, "called thread=0x{:X}", thread_handle);
@@ -917,6 +951,18 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
     }
 }
 
+static ResultCode GetInfo32(Core::System& system, u32* result_low, u32* result_high, u32 sub_id_low,
+                            u32 info_id, u32 handle, u32 sub_id_high) {
+    const u64 sub_id{static_cast<u64>(sub_id_low | (static_cast<u64>(sub_id_high) << 32))};
+    u64 res_value{};
+
+    const ResultCode result{GetInfo(system, &res_value, info_id, handle, sub_id)};
+    *result_high = static_cast<u32>(res_value >> 32);
+    *result_low = static_cast<u32>(res_value & std::numeric_limits<u32>::max());
+
+    return result;
+}
+
 /// Maps memory at a desired address
 static ResultCode MapPhysicalMemory(Core::System& system, VAddr addr, u64 size) {
     LOG_DEBUG(Kernel_SVC, "called, addr=0x{:016X}, size=0x{:X}", addr, size);
@@ -1058,7 +1104,7 @@ static ResultCode GetThreadContext(Core::System& system, VAddr thread_context, H
         return ERR_BUSY;
     }
 
-    Core::ARM_Interface::ThreadContext ctx = thread->GetContext();
+    Core::ARM_Interface::ThreadContext64 ctx = thread->GetContext64();
     // Mask away mode bits, interrupt bits, IL bit, and other reserved bits.
     ctx.pstate &= 0xFF0FFE20;
 
@@ -1088,6 +1134,10 @@ static ResultCode GetThreadPriority(Core::System& system, u32* priority, Handle
     return RESULT_SUCCESS;
 }
 
+static ResultCode GetThreadPriority32(Core::System& system, u32* priority, Handle handle) {
+    return GetThreadPriority(system, priority, handle);
+}
+
 /// Sets the priority for the specified thread
 static ResultCode SetThreadPriority(Core::System& system, Handle handle, u32 priority) {
     LOG_TRACE(Kernel_SVC, "called");
@@ -1259,6 +1309,11 @@ static ResultCode QueryMemory(Core::System& system, VAddr memory_info_address,
                               query_address);
 }
 
+static ResultCode QueryMemory32(Core::System& system, u32 memory_info_address,
+                                u32 page_info_address, u32 query_address) {
+    return QueryMemory(system, memory_info_address, page_info_address, query_address);
+}
+
 static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_handle, u64 dst_address,
                                        u64 src_address, u64 size) {
     LOG_DEBUG(Kernel_SVC,
@@ -1675,6 +1730,10 @@ static void SignalProcessWideKey(Core::System& system, VAddr condition_variable_
     }
 }
 
+static void SignalProcessWideKey32(Core::System& system, u32 condition_variable_addr, s32 target) {
+    SignalProcessWideKey(system, condition_variable_addr, target);
+}
+
 // Wait for an address (via Address Arbiter)
 static ResultCode WaitForAddress(Core::System& system, VAddr address, u32 type, s32 value,
                                  s64 timeout) {
@@ -1760,6 +1819,10 @@ static ResultCode CloseHandle(Core::System& system, Handle handle) {
     return handle_table.Close(handle);
 }
 
+static ResultCode CloseHandle32(Core::System& system, Handle handle) {
+    return CloseHandle(system, handle);
+}
+
 /// Clears the signaled state of an event or process.
 static ResultCode ResetSignal(Core::System& system, Handle handle) {
     LOG_DEBUG(Kernel_SVC, "called handle 0x{:08X}", handle);
@@ -2317,69 +2380,196 @@ struct FunctionDef {
 };
 } // namespace
 
-static const FunctionDef SVC_Table[] = {
+static const FunctionDef SVC_Table_32[] = {
     {0x00, nullptr, "Unknown"},
-    {0x01, SvcWrap<SetHeapSize>, "SetHeapSize"},
-    {0x02, SvcWrap<SetMemoryPermission>, "SetMemoryPermission"},
-    {0x03, SvcWrap<SetMemoryAttribute>, "SetMemoryAttribute"},
-    {0x04, SvcWrap<MapMemory>, "MapMemory"},
-    {0x05, SvcWrap<UnmapMemory>, "UnmapMemory"},
-    {0x06, SvcWrap<QueryMemory>, "QueryMemory"},
-    {0x07, SvcWrap<ExitProcess>, "ExitProcess"},
-    {0x08, SvcWrap<CreateThread>, "CreateThread"},
-    {0x09, SvcWrap<StartThread>, "StartThread"},
-    {0x0A, SvcWrap<ExitThread>, "ExitThread"},
-    {0x0B, SvcWrap<SleepThread>, "SleepThread"},
-    {0x0C, SvcWrap<GetThreadPriority>, "GetThreadPriority"},
-    {0x0D, SvcWrap<SetThreadPriority>, "SetThreadPriority"},
-    {0x0E, SvcWrap<GetThreadCoreMask>, "GetThreadCoreMask"},
-    {0x0F, SvcWrap<SetThreadCoreMask>, "SetThreadCoreMask"},
-    {0x10, SvcWrap<GetCurrentProcessorNumber>, "GetCurrentProcessorNumber"},
-    {0x11, SvcWrap<SignalEvent>, "SignalEvent"},
-    {0x12, SvcWrap<ClearEvent>, "ClearEvent"},
-    {0x13, SvcWrap<MapSharedMemory>, "MapSharedMemory"},
-    {0x14, SvcWrap<UnmapSharedMemory>, "UnmapSharedMemory"},
-    {0x15, SvcWrap<CreateTransferMemory>, "CreateTransferMemory"},
-    {0x16, SvcWrap<CloseHandle>, "CloseHandle"},
-    {0x17, SvcWrap<ResetSignal>, "ResetSignal"},
-    {0x18, SvcWrap<WaitSynchronization>, "WaitSynchronization"},
-    {0x19, SvcWrap<CancelSynchronization>, "CancelSynchronization"},
-    {0x1A, SvcWrap<ArbitrateLock>, "ArbitrateLock"},
-    {0x1B, SvcWrap<ArbitrateUnlock>, "ArbitrateUnlock"},
-    {0x1C, SvcWrap<WaitProcessWideKeyAtomic>, "WaitProcessWideKeyAtomic"},
-    {0x1D, SvcWrap<SignalProcessWideKey>, "SignalProcessWideKey"},
-    {0x1E, SvcWrap<GetSystemTick>, "GetSystemTick"},
-    {0x1F, SvcWrap<ConnectToNamedPort>, "ConnectToNamedPort"},
+    {0x01, SvcWrap32<SetHeapSize32>, "SetHeapSize32"},
+    {0x02, nullptr, "Unknown"},
+    {0x03, nullptr, "SetMemoryAttribute32"},
+    {0x04, nullptr, "MapMemory32"},
+    {0x05, nullptr, "UnmapMemory32"},
+    {0x06, SvcWrap32<QueryMemory32>, "QueryMemory32"},
+    {0x07, nullptr, "ExitProcess32"},
+    {0x08, nullptr, "CreateThread32"},
+    {0x09, nullptr, "StartThread32"},
+    {0x0a, nullptr, "ExitThread32"},
+    {0x0b, nullptr, "SleepThread32"},
+    {0x0c, SvcWrap32<GetThreadPriority32>, "GetThreadPriority32"},
+    {0x0d, nullptr, "SetThreadPriority32"},
+    {0x0e, nullptr, "GetThreadCoreMask32"},
+    {0x0f, nullptr, "SetThreadCoreMask32"},
+    {0x10, nullptr, "GetCurrentProcessorNumber32"},
+    {0x11, nullptr, "SignalEvent32"},
+    {0x12, nullptr, "ClearEvent32"},
+    {0x13, nullptr, "MapSharedMemory32"},
+    {0x14, nullptr, "UnmapSharedMemory32"},
+    {0x15, nullptr, "CreateTransferMemory32"},
+    {0x16, SvcWrap32<CloseHandle32>, "CloseHandle32"},
+    {0x17, nullptr, "ResetSignal32"},
+    {0x18, SvcWrap32<WaitSynchronization32>, "WaitSynchronization32"},
+    {0x19, nullptr, "CancelSynchronization32"},
+    {0x1a, nullptr, "ArbitrateLock32"},
+    {0x1b, nullptr, "ArbitrateUnlock32"},
+    {0x1c, nullptr, "WaitProcessWideKeyAtomic32"},
+    {0x1d, SvcWrap32<SignalProcessWideKey32>, "SignalProcessWideKey32"},
+    {0x1e, nullptr, "GetSystemTick32"},
+    {0x1f, SvcWrap32<ConnectToNamedPort32>, "ConnectToNamedPort32"},
+    {0x20, nullptr, "Unknown"},
+    {0x21, SvcWrap32<SendSyncRequest32>, "SendSyncRequest32"},
+    {0x22, nullptr, "SendSyncRequestWithUserBuffer32"},
+    {0x23, nullptr, "Unknown"},
+    {0x24, nullptr, "GetProcessId32"},
+    {0x25, SvcWrap32<GetThreadId32>, "GetThreadId32"},
+    {0x26, nullptr, "Break32"},
+    {0x27, nullptr, "OutputDebugString32"},
+    {0x28, nullptr, "Unknown"},
+    {0x29, SvcWrap32<GetInfo32>, "GetInfo32"},
+    {0x2a, nullptr, "Unknown"},
+    {0x2b, nullptr, "Unknown"},
+    {0x2c, nullptr, "MapPhysicalMemory32"},
+    {0x2d, nullptr, "UnmapPhysicalMemory32"},
+    {0x2e, nullptr, "Unknown"},
+    {0x2f, nullptr, "Unknown"},
+    {0x30, nullptr, "Unknown"},
+    {0x31, nullptr, "Unknown"},
+    {0x32, nullptr, "SetThreadActivity32"},
+    {0x33, nullptr, "GetThreadContext32"},
+    {0x34, nullptr, "WaitForAddress32"},
+    {0x35, nullptr, "SignalToAddress32"},
+    {0x36, nullptr, "Unknown"},
+    {0x37, nullptr, "Unknown"},
+    {0x38, nullptr, "Unknown"},
+    {0x39, nullptr, "Unknown"},
+    {0x3a, nullptr, "Unknown"},
+    {0x3b, nullptr, "Unknown"},
+    {0x3c, nullptr, "Unknown"},
+    {0x3d, nullptr, "Unknown"},
+    {0x3e, nullptr, "Unknown"},
+    {0x3f, nullptr, "Unknown"},
+    {0x40, nullptr, "CreateSession32"},
+    {0x41, nullptr, "AcceptSession32"},
+    {0x42, nullptr, "Unknown"},
+    {0x43, nullptr, "ReplyAndReceive32"},
+    {0x44, nullptr, "Unknown"},
+    {0x45, nullptr, "CreateEvent32"},
+    {0x46, nullptr, "Unknown"},
+    {0x47, nullptr, "Unknown"},
+    {0x48, nullptr, "Unknown"},
+    {0x49, nullptr, "Unknown"},
+    {0x4a, nullptr, "Unknown"},
+    {0x4b, nullptr, "Unknown"},
+    {0x4c, nullptr, "Unknown"},
+    {0x4d, nullptr, "Unknown"},
+    {0x4e, nullptr, "Unknown"},
+    {0x4f, nullptr, "Unknown"},
+    {0x50, nullptr, "Unknown"},
+    {0x51, nullptr, "Unknown"},
+    {0x52, nullptr, "Unknown"},
+    {0x53, nullptr, "Unknown"},
+    {0x54, nullptr, "Unknown"},
+    {0x55, nullptr, "Unknown"},
+    {0x56, nullptr, "Unknown"},
+    {0x57, nullptr, "Unknown"},
+    {0x58, nullptr, "Unknown"},
+    {0x59, nullptr, "Unknown"},
+    {0x5a, nullptr, "Unknown"},
+    {0x5b, nullptr, "Unknown"},
+    {0x5c, nullptr, "Unknown"},
+    {0x5d, nullptr, "Unknown"},
+    {0x5e, nullptr, "Unknown"},
+    {0x5F, nullptr, "FlushProcessDataCache32"},
+    {0x60, nullptr, "Unknown"},
+    {0x61, nullptr, "Unknown"},
+    {0x62, nullptr, "Unknown"},
+    {0x63, nullptr, "Unknown"},
+    {0x64, nullptr, "Unknown"},
+    {0x65, nullptr, "GetProcessList32"},
+    {0x66, nullptr, "Unknown"},
+    {0x67, nullptr, "Unknown"},
+    {0x68, nullptr, "Unknown"},
+    {0x69, nullptr, "Unknown"},
+    {0x6A, nullptr, "Unknown"},
+    {0x6B, nullptr, "Unknown"},
+    {0x6C, nullptr, "Unknown"},
+    {0x6D, nullptr, "Unknown"},
+    {0x6E, nullptr, "Unknown"},
+    {0x6f, nullptr, "GetSystemInfo32"},
+    {0x70, nullptr, "CreatePort32"},
+    {0x71, nullptr, "ManageNamedPort32"},
+    {0x72, nullptr, "ConnectToPort32"},
+    {0x73, nullptr, "SetProcessMemoryPermission32"},
+    {0x74, nullptr, "Unknown"},
+    {0x75, nullptr, "Unknown"},
+    {0x76, nullptr, "Unknown"},
+    {0x77, nullptr, "MapProcessCodeMemory32"},
+    {0x78, nullptr, "UnmapProcessCodeMemory32"},
+    {0x79, nullptr, "Unknown"},
+    {0x7A, nullptr, "Unknown"},
+    {0x7B, nullptr, "TerminateProcess32"},
+};
+
+static const FunctionDef SVC_Table_64[] = {
+    {0x00, nullptr, "Unknown"},
+    {0x01, SvcWrap64<SetHeapSize>, "SetHeapSize"},
+    {0x02, SvcWrap64<SetMemoryPermission>, "SetMemoryPermission"},
+    {0x03, SvcWrap64<SetMemoryAttribute>, "SetMemoryAttribute"},
+    {0x04, SvcWrap64<MapMemory>, "MapMemory"},
+    {0x05, SvcWrap64<UnmapMemory>, "UnmapMemory"},
+    {0x06, SvcWrap64<QueryMemory>, "QueryMemory"},
+    {0x07, SvcWrap64<ExitProcess>, "ExitProcess"},
+    {0x08, SvcWrap64<CreateThread>, "CreateThread"},
+    {0x09, SvcWrap64<StartThread>, "StartThread"},
+    {0x0A, SvcWrap64<ExitThread>, "ExitThread"},
+    {0x0B, SvcWrap64<SleepThread>, "SleepThread"},
+    {0x0C, SvcWrap64<GetThreadPriority>, "GetThreadPriority"},
+    {0x0D, SvcWrap64<SetThreadPriority>, "SetThreadPriority"},
+    {0x0E, SvcWrap64<GetThreadCoreMask>, "GetThreadCoreMask"},
+    {0x0F, SvcWrap64<SetThreadCoreMask>, "SetThreadCoreMask"},
+    {0x10, SvcWrap64<GetCurrentProcessorNumber>, "GetCurrentProcessorNumber"},
+    {0x11, SvcWrap64<SignalEvent>, "SignalEvent"},
+    {0x12, SvcWrap64<ClearEvent>, "ClearEvent"},
+    {0x13, SvcWrap64<MapSharedMemory>, "MapSharedMemory"},
+    {0x14, SvcWrap64<UnmapSharedMemory>, "UnmapSharedMemory"},
+    {0x15, SvcWrap64<CreateTransferMemory>, "CreateTransferMemory"},
+    {0x16, SvcWrap64<CloseHandle>, "CloseHandle"},
+    {0x17, SvcWrap64<ResetSignal>, "ResetSignal"},
+    {0x18, SvcWrap64<WaitSynchronization>, "WaitSynchronization"},
+    {0x19, SvcWrap64<CancelSynchronization>, "CancelSynchronization"},
+    {0x1A, SvcWrap64<ArbitrateLock>, "ArbitrateLock"},
+    {0x1B, SvcWrap64<ArbitrateUnlock>, "ArbitrateUnlock"},
+    {0x1C, SvcWrap64<WaitProcessWideKeyAtomic>, "WaitProcessWideKeyAtomic"},
+    {0x1D, SvcWrap64<SignalProcessWideKey>, "SignalProcessWideKey"},
+    {0x1E, SvcWrap64<GetSystemTick>, "GetSystemTick"},
+    {0x1F, SvcWrap64<ConnectToNamedPort>, "ConnectToNamedPort"},
     {0x20, nullptr, "SendSyncRequestLight"},
-    {0x21, SvcWrap<SendSyncRequest>, "SendSyncRequest"},
+    {0x21, SvcWrap64<SendSyncRequest>, "SendSyncRequest"},
     {0x22, nullptr, "SendSyncRequestWithUserBuffer"},
     {0x23, nullptr, "SendAsyncRequestWithUserBuffer"},
-    {0x24, SvcWrap<GetProcessId>, "GetProcessId"},
-    {0x25, SvcWrap<GetThreadId>, "GetThreadId"},
-    {0x26, SvcWrap<Break>, "Break"},
-    {0x27, SvcWrap<OutputDebugString>, "OutputDebugString"},
+    {0x24, SvcWrap64<GetProcessId>, "GetProcessId"},
+    {0x25, SvcWrap64<GetThreadId>, "GetThreadId"},
+    {0x26, SvcWrap64<Break>, "Break"},
+    {0x27, SvcWrap64<OutputDebugString>, "OutputDebugString"},
     {0x28, nullptr, "ReturnFromException"},
-    {0x29, SvcWrap<GetInfo>, "GetInfo"},
+    {0x29, SvcWrap64<GetInfo>, "GetInfo"},
     {0x2A, nullptr, "FlushEntireDataCache"},
     {0x2B, nullptr, "FlushDataCache"},
-    {0x2C, SvcWrap<MapPhysicalMemory>, "MapPhysicalMemory"},
-    {0x2D, SvcWrap<UnmapPhysicalMemory>, "UnmapPhysicalMemory"},
+    {0x2C, SvcWrap64<MapPhysicalMemory>, "MapPhysicalMemory"},
+    {0x2D, SvcWrap64<UnmapPhysicalMemory>, "UnmapPhysicalMemory"},
     {0x2E, nullptr, "GetFutureThreadInfo"},
     {0x2F, nullptr, "GetLastThreadInfo"},
-    {0x30, SvcWrap<GetResourceLimitLimitValue>, "GetResourceLimitLimitValue"},
-    {0x31, SvcWrap<GetResourceLimitCurrentValue>, "GetResourceLimitCurrentValue"},
-    {0x32, SvcWrap<SetThreadActivity>, "SetThreadActivity"},
-    {0x33, SvcWrap<GetThreadContext>, "GetThreadContext"},
-    {0x34, SvcWrap<WaitForAddress>, "WaitForAddress"},
-    {0x35, SvcWrap<SignalToAddress>, "SignalToAddress"},
+    {0x30, SvcWrap64<GetResourceLimitLimitValue>, "GetResourceLimitLimitValue"},
+    {0x31, SvcWrap64<GetResourceLimitCurrentValue>, "GetResourceLimitCurrentValue"},
+    {0x32, SvcWrap64<SetThreadActivity>, "SetThreadActivity"},
+    {0x33, SvcWrap64<GetThreadContext>, "GetThreadContext"},
+    {0x34, SvcWrap64<WaitForAddress>, "WaitForAddress"},
+    {0x35, SvcWrap64<SignalToAddress>, "SignalToAddress"},
     {0x36, nullptr, "SynchronizePreemptionState"},
     {0x37, nullptr, "Unknown"},
     {0x38, nullptr, "Unknown"},
     {0x39, nullptr, "Unknown"},
     {0x3A, nullptr, "Unknown"},
     {0x3B, nullptr, "Unknown"},
-    {0x3C, SvcWrap<KernelDebug>, "KernelDebug"},
-    {0x3D, SvcWrap<ChangeKernelTraceState>, "ChangeKernelTraceState"},
+    {0x3C, SvcWrap64<KernelDebug>, "KernelDebug"},
+    {0x3D, SvcWrap64<ChangeKernelTraceState>, "ChangeKernelTraceState"},
     {0x3E, nullptr, "Unknown"},
     {0x3F, nullptr, "Unknown"},
     {0x40, nullptr, "CreateSession"},
@@ -2387,7 +2577,7 @@ static const FunctionDef SVC_Table[] = {
     {0x42, nullptr, "ReplyAndReceiveLight"},
     {0x43, nullptr, "ReplyAndReceive"},
     {0x44, nullptr, "ReplyAndReceiveWithUserBuffer"},
-    {0x45, SvcWrap<CreateEvent>, "CreateEvent"},
+    {0x45, SvcWrap64<CreateEvent>, "CreateEvent"},
     {0x46, nullptr, "Unknown"},
     {0x47, nullptr, "Unknown"},
     {0x48, nullptr, "MapPhysicalMemoryUnsafe"},
@@ -2398,9 +2588,9 @@ static const FunctionDef SVC_Table[] = {
     {0x4D, nullptr, "SleepSystem"},
     {0x4E, nullptr, "ReadWriteRegister"},
     {0x4F, nullptr, "SetProcessActivity"},
-    {0x50, SvcWrap<CreateSharedMemory>, "CreateSharedMemory"},
-    {0x51, SvcWrap<MapTransferMemory>, "MapTransferMemory"},
-    {0x52, SvcWrap<UnmapTransferMemory>, "UnmapTransferMemory"},
+    {0x50, SvcWrap64<CreateSharedMemory>, "CreateSharedMemory"},
+    {0x51, SvcWrap64<MapTransferMemory>, "MapTransferMemory"},
+    {0x52, SvcWrap64<UnmapTransferMemory>, "UnmapTransferMemory"},
     {0x53, nullptr, "CreateInterruptEvent"},
     {0x54, nullptr, "QueryPhysicalAddress"},
     {0x55, nullptr, "QueryIoMapping"},
@@ -2419,8 +2609,8 @@ static const FunctionDef SVC_Table[] = {
     {0x62, nullptr, "TerminateDebugProcess"},
     {0x63, nullptr, "GetDebugEvent"},
     {0x64, nullptr, "ContinueDebugEvent"},
-    {0x65, SvcWrap<GetProcessList>, "GetProcessList"},
-    {0x66, SvcWrap<GetThreadList>, "GetThreadList"},
+    {0x65, SvcWrap64<GetProcessList>, "GetProcessList"},
+    {0x66, SvcWrap64<GetThreadList>, "GetThreadList"},
     {0x67, nullptr, "GetDebugThreadContext"},
     {0x68, nullptr, "SetDebugThreadContext"},
     {0x69, nullptr, "QueryDebugProcessMemory"},
@@ -2436,24 +2626,32 @@ static const FunctionDef SVC_Table[] = {
     {0x73, nullptr, "SetProcessMemoryPermission"},
     {0x74, nullptr, "MapProcessMemory"},
     {0x75, nullptr, "UnmapProcessMemory"},
-    {0x76, SvcWrap<QueryProcessMemory>, "QueryProcessMemory"},
-    {0x77, SvcWrap<MapProcessCodeMemory>, "MapProcessCodeMemory"},
-    {0x78, SvcWrap<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
+    {0x76, SvcWrap64<QueryProcessMemory>, "QueryProcessMemory"},
+    {0x77, SvcWrap64<MapProcessCodeMemory>, "MapProcessCodeMemory"},
+    {0x78, SvcWrap64<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
     {0x79, nullptr, "CreateProcess"},
     {0x7A, nullptr, "StartProcess"},
     {0x7B, nullptr, "TerminateProcess"},
-    {0x7C, SvcWrap<GetProcessInfo>, "GetProcessInfo"},
-    {0x7D, SvcWrap<CreateResourceLimit>, "CreateResourceLimit"},
-    {0x7E, SvcWrap<SetResourceLimitLimitValue>, "SetResourceLimitLimitValue"},
+    {0x7C, SvcWrap64<GetProcessInfo>, "GetProcessInfo"},
+    {0x7D, SvcWrap64<CreateResourceLimit>, "CreateResourceLimit"},
+    {0x7E, SvcWrap64<SetResourceLimitLimitValue>, "SetResourceLimitLimitValue"},
     {0x7F, nullptr, "CallSecureMonitor"},
 };
 
-static const FunctionDef* GetSVCInfo(u32 func_num) {
-    if (func_num >= std::size(SVC_Table)) {
+static const FunctionDef* GetSVCInfo32(u32 func_num) {
+    if (func_num >= std::size(SVC_Table_32)) {
+        LOG_ERROR(Kernel_SVC, "Unknown svc=0x{:02X}", func_num);
+        return nullptr;
+    }
+    return &SVC_Table_32[func_num];
+}
+
+static const FunctionDef* GetSVCInfo64(u32 func_num) {
+    if (func_num >= std::size(SVC_Table_64)) {
         LOG_ERROR(Kernel_SVC, "Unknown svc=0x{:02X}", func_num);
         return nullptr;
     }
-    return &SVC_Table[func_num];
+    return &SVC_Table_64[func_num];
 }
 
 MICROPROFILE_DEFINE(Kernel_SVC, "Kernel", "SVC", MP_RGB(70, 200, 70));
@@ -2464,7 +2662,8 @@ void CallSVC(Core::System& system, u32 immediate) {
     // Lock the global kernel mutex when we enter the kernel HLE.
     std::lock_guard lock{HLE::g_hle_lock};
 
-    const FunctionDef* info = GetSVCInfo(immediate);
+    const FunctionDef* info = system.CurrentProcess()->Is64BitProcess() ? GetSVCInfo64(immediate)
+                                                                        : GetSVCInfo32(immediate);
     if (info) {
         if (info->func) {
             info->func(system);
diff --git a/src/core/hle/kernel/svc_wrap.h b/src/core/hle/kernel/svc_wrap.h
index 29a2cfa9d..7d735e3fa 100644
--- a/src/core/hle/kernel/svc_wrap.h
+++ b/src/core/hle/kernel/svc_wrap.h
@@ -15,6 +15,10 @@ static inline u64 Param(const Core::System& system, int n) {
     return system.CurrentArmInterface().GetReg(n);
 }
 
+static inline u32 Param32(const Core::System& system, int n) {
+    return static_cast<u32>(system.CurrentArmInterface().GetReg(n));
+}
+
 /**
  * HLE a function return from the current ARM userland process
  * @param system System context
@@ -24,40 +28,44 @@ static inline void FuncReturn(Core::System& system, u64 result) {
     system.CurrentArmInterface().SetReg(0, result);
 }
 
+static inline void FuncReturn32(Core::System& system, u32 result) {
+    system.CurrentArmInterface().SetReg(0, (u64)result);
+}
+
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Function wrappers that return type ResultCode
 
 template <ResultCode func(Core::System&, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0)).raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1)).raw);
 }
 
 template <ResultCode func(Core::System&, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(
         system,
         func(system, static_cast<u32>(Param(system, 0)), static_cast<u32>(Param(system, 1))).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                             Param(system, 2), Param(system, 3))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param = 0;
     const u32 retval = func(system, &param).raw;
     system.CurrentArmInterface().SetReg(1, param);
@@ -65,7 +73,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
     system.CurrentArmInterface().SetReg(1, param_1);
@@ -73,7 +81,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u32*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     u32 param_2 = 0;
     const u32 retval = func(system, &param_1, &param_2).raw;
@@ -86,7 +94,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1)).raw;
     system.CurrentArmInterface().SetReg(1, param_1);
@@ -94,7 +102,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval =
         func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2))).raw;
@@ -104,7 +112,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64*, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
 
@@ -113,12 +121,12 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1))).raw);
 }
 
 template <ResultCode func(Core::System&, u64*, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1)).raw;
 
@@ -127,7 +135,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64*, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1)),
                             static_cast<u32>(Param(system, 2)))
@@ -138,19 +146,19 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1)).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)),
                             static_cast<u32>(Param(system, 1)), Param(system, 2))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32, u32*, u64*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     u64 param_2 = 0;
     const ResultCode retval = func(system, static_cast<u32>(Param(system, 2)), &param_1, &param_2);
@@ -161,54 +169,54 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                             static_cast<u32>(Param(system, 2)), static_cast<u32>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                             static_cast<u32>(Param(system, 2)), Param(system, 3))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                             static_cast<u32>(Param(system, 2)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1), Param(system, 2)).raw);
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(
         system,
         func(system, Param(system, 0), Param(system, 1), static_cast<u32>(Param(system, 2))).raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                             Param(system, 2), static_cast<u32>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u32, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(
         system,
         func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2)).raw);
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u64, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
                             static_cast<s64>(Param(system, 3)))
@@ -219,14 +227,14 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u64, u32, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                             static_cast<u32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64*, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u64 param_1 = 0;
     const u32 retval =
         func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3)).raw;
@@ -236,7 +244,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u64, u64, u32, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3),
                             static_cast<u32>(Param(system, 4)), static_cast<s32>(Param(system, 5)))
@@ -247,7 +255,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u32*, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2),
                             static_cast<u32>(Param(system, 3)))
@@ -258,7 +266,7 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, Handle*, u64, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     u32 param_1 = 0;
     const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
                             static_cast<u32>(Param(system, 3)))
@@ -269,14 +277,14 @@ void SvcWrap(Core::System& system) {
 }
 
 template <ResultCode func(Core::System&, u64, u32, s32, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
                             static_cast<s32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
                            .raw);
 }
 
 template <ResultCode func(Core::System&, u64, u32, s32, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
                             static_cast<s32>(Param(system, 2)), static_cast<s32>(Param(system, 3)))
                            .raw);
@@ -286,7 +294,7 @@ void SvcWrap(Core::System& system) {
 // Function wrappers that return type u32
 
 template <u32 func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system));
 }
 
@@ -294,7 +302,7 @@ void SvcWrap(Core::System& system) {
 // Function wrappers that return type u64
 
 template <u64 func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     FuncReturn(system, func(system));
 }
 
@@ -302,44 +310,110 @@ void SvcWrap(Core::System& system) {
 /// Function wrappers that return type void
 
 template <void func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system);
 }
 
 template <void func(Core::System&, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<u32>(Param(system, 0)));
 }
 
 template <void func(Core::System&, u32, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2),
          Param(system, 3));
 }
 
 template <void func(Core::System&, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<s64>(Param(system, 0)));
 }
 
 template <void func(Core::System&, u64, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, Param(system, 0), static_cast<s32>(Param(system, 1)));
 }
 
 template <void func(Core::System&, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, Param(system, 0), Param(system, 1));
 }
 
 template <void func(Core::System&, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, Param(system, 0), Param(system, 1), Param(system, 2));
 }
 
 template <void func(Core::System&, u32, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
     func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2));
 }
 
+// Used by QueryMemory32
+template <ResultCode func(Core::System&, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn32(system,
+                 func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2)).raw);
+}
+
+// Used by GetInfo32
+template <ResultCode func(Core::System&, u32*, u32*, u32, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 0), Param32(system, 1),
+                            Param32(system, 2), Param32(system, 3))
+                           .raw;
+
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+
+// Used by GetThreadPriority32, ConnectToNamedPort32
+template <ResultCode func(Core::System&, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
+// Used by GetThreadId32
+template <ResultCode func(Core::System&, u32*, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+
+// Used by SignalProcessWideKey32
+template <void func(Core::System&, u32, s32)>
+void SvcWrap32(Core::System& system) {
+    func(system, static_cast<u32>(Param(system, 0)), static_cast<s32>(Param(system, 1)));
+}
+
+// Used by SendSyncRequest32
+template <ResultCode func(Core::System&, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
+}
+
+// Used by WaitSynchronization32
+template <ResultCode func(Core::System&, u32, u32, s32, u32, Handle*)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2),
+                            Param32(system, 3), &param_1)
+                           .raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index bf850e0b2..83e956036 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -133,15 +133,16 @@ void Thread::CancelWait() {
     ResumeFromWait();
 }
 
-/**
- * Resets a thread context, making it ready to be scheduled and run by the CPU
- * @param context Thread context to reset
- * @param stack_top Address of the top of the stack
- * @param entry_point Address of entry point for execution
- * @param arg User argument for thread
- */
-static void ResetThreadContext(Core::ARM_Interface::ThreadContext& context, VAddr stack_top,
-                               VAddr entry_point, u64 arg) {
+static void ResetThreadContext32(Core::ARM_Interface::ThreadContext32& context, u32 stack_top,
+                                 u32 entry_point, u32 arg) {
+    context = {};
+    context.cpu_registers[0] = arg;
+    context.cpu_registers[15] = entry_point;
+    context.cpu_registers[13] = stack_top;
+}
+
+static void ResetThreadContext64(Core::ARM_Interface::ThreadContext64& context, VAddr stack_top,
+                                 VAddr entry_point, u64 arg) {
     context = {};
     context.cpu_registers[0] = arg;
     context.pc = entry_point;
@@ -198,9 +199,9 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(KernelCore& kernel, std::strin
 
     thread->owner_process->RegisterThread(thread.get());
 
-    // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
-    // to initialize the context
-    ResetThreadContext(thread->context, stack_top, entry_point, arg);
+    ResetThreadContext32(thread->context_32, static_cast<u32>(stack_top),
+                         static_cast<u32>(entry_point), static_cast<u32>(arg));
+    ResetThreadContext64(thread->context_64, stack_top, entry_point, arg);
 
     return MakeResult<std::shared_ptr<Thread>>(std::move(thread));
 }
@@ -213,11 +214,13 @@ void Thread::SetPriority(u32 priority) {
 }
 
 void Thread::SetWaitSynchronizationResult(ResultCode result) {
-    context.cpu_registers[0] = result.raw;
+    context_32.cpu_registers[0] = result.raw;
+    context_64.cpu_registers[0] = result.raw;
 }
 
 void Thread::SetWaitSynchronizationOutput(s32 output) {
-    context.cpu_registers[1] = output;
+    context_32.cpu_registers[1] = output;
+    context_64.cpu_registers[1] = output;
 }
 
 s32 Thread::GetSynchronizationObjectIndex(std::shared_ptr<SynchronizationObject> object) const {
diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h
index 129e7858a..23fdef8a4 100644
--- a/src/core/hle/kernel/thread.h
+++ b/src/core/hle/kernel/thread.h
@@ -102,7 +102,8 @@ public:
 
     using MutexWaitingThreads = std::vector<std::shared_ptr<Thread>>;
 
-    using ThreadContext = Core::ARM_Interface::ThreadContext;
+    using ThreadContext32 = Core::ARM_Interface::ThreadContext32;
+    using ThreadContext64 = Core::ARM_Interface::ThreadContext64;
 
     using ThreadSynchronizationObjects = std::vector<std::shared_ptr<SynchronizationObject>>;
 
@@ -273,12 +274,20 @@ public:
         return status == ThreadStatus::WaitSynch;
     }
 
-    ThreadContext& GetContext() {
-        return context;
+    ThreadContext32& GetContext32() {
+        return context_32;
     }
 
-    const ThreadContext& GetContext() const {
-        return context;
+    const ThreadContext32& GetContext32() const {
+        return context_32;
+    }
+
+    ThreadContext64& GetContext64() {
+        return context_64;
+    }
+
+    const ThreadContext64& GetContext64() const {
+        return context_64;
     }
 
     ThreadStatus GetStatus() const {
@@ -466,7 +475,8 @@ private:
     void AdjustSchedulingOnPriority(u32 old_priority);
     void AdjustSchedulingOnAffinity(u64 old_affinity_mask, s32 old_core);
 
-    Core::ARM_Interface::ThreadContext context{};
+    ThreadContext32 context_32{};
+    ThreadContext64 context_64{};
 
     u64 thread_id = 0;
 
diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp
index cc978713b..d1bf13c89 100644
--- a/src/core/hle/service/am/am.cpp
+++ b/src/core/hle/service/am/am.cpp
@@ -607,7 +607,7 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system,
         {40, nullptr, "GetCradleFwVersion"},
         {50, nullptr, "IsVrModeEnabled"},
         {51, nullptr, "SetVrModeEnabled"},
-        {52, nullptr, "SwitchLcdBacklight"},
+        {52, &ICommonStateGetter::SetLcdBacklighOffEnabled, "SetLcdBacklighOffEnabled"},
         {53, nullptr, "BeginVrModeEx"},
         {54, nullptr, "EndVrModeEx"},
         {55, nullptr, "IsInControllerFirmwareUpdateSection"},
@@ -636,7 +636,6 @@ void ICommonStateGetter::GetBootMode(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-
     rb.Push<u8>(static_cast<u8>(Service::PM::SystemBootMode::Normal)); // Normal boot mode
 }
 
@@ -660,6 +659,7 @@ void ICommonStateGetter::ReceiveMessage(Kernel::HLERequestContext& ctx) {
         rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
         return;
     }
+
     rb.Push(RESULT_SUCCESS);
     rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
 }
@@ -672,6 +672,17 @@ void ICommonStateGetter::GetCurrentFocusState(Kernel::HLERequestContext& ctx) {
     rb.Push(static_cast<u8>(FocusState::InFocus));
 }
 
+void ICommonStateGetter::SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto is_lcd_backlight_off_enabled = rp.Pop<bool>();
+
+    LOG_WARNING(Service_AM, "(STUBBED) called. is_lcd_backlight_off_enabled={}",
+                is_lcd_backlight_off_enabled);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void ICommonStateGetter::GetDefaultDisplayResolutionChangeEvent(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called");
 
diff --git a/src/core/hle/service/am/am.h b/src/core/hle/service/am/am.h
index 0b9a4332d..0843de781 100644
--- a/src/core/hle/service/am/am.h
+++ b/src/core/hle/service/am/am.h
@@ -182,6 +182,7 @@ private:
     void GetOperationMode(Kernel::HLERequestContext& ctx);
     void GetPerformanceMode(Kernel::HLERequestContext& ctx);
     void GetBootMode(Kernel::HLERequestContext& ctx);
+    void SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx);
     void GetDefaultDisplayResolution(Kernel::HLERequestContext& ctx);
     void SetCpuBoostMode(Kernel::HLERequestContext& ctx);
 
diff --git a/src/core/hle/service/am/applets/web_browser.cpp b/src/core/hle/service/am/applets/web_browser.cpp
index 12443c910..9f30e167d 100644
--- a/src/core/hle/service/am/applets/web_browser.cpp
+++ b/src/core/hle/service/am/applets/web_browser.cpp
@@ -254,6 +254,12 @@ void WebBrowser::Execute() {
 
     if (status != RESULT_SUCCESS) {
         complete = true;
+
+        // This is a workaround in order not to softlock yuzu when an error happens during the
+        // webapplet init. In order to avoid an svcBreak, the status is set to RESULT_SUCCESS
+        Finalize();
+        status = RESULT_SUCCESS;
+
         return;
     }
 
diff --git a/src/core/hle/service/hid/controllers/npad.cpp b/src/core/hle/service/hid/controllers/npad.cpp
index 15c09f04c..c1e32b28c 100644
--- a/src/core/hle/service/hid/controllers/npad.cpp
+++ b/src/core/hle/service/hid/controllers/npad.cpp
@@ -287,13 +287,13 @@ void Controller_NPad::RequestPadStateUpdate(u32 npad_id) {
         analog_state[static_cast<std::size_t>(JoystickId::Joystick_Left)]->GetAnalogDirectionStatus(
             Input::AnalogDirection::DOWN));
 
-    pad_state.r_stick_up.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
-                                    ->GetAnalogDirectionStatus(Input::AnalogDirection::RIGHT));
-    pad_state.r_stick_left.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
-                                      ->GetAnalogDirectionStatus(Input::AnalogDirection::LEFT));
     pad_state.r_stick_right.Assign(
         analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
-            ->GetAnalogDirectionStatus(Input::AnalogDirection::UP));
+            ->GetAnalogDirectionStatus(Input::AnalogDirection::RIGHT));
+    pad_state.r_stick_left.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
+                                      ->GetAnalogDirectionStatus(Input::AnalogDirection::LEFT));
+    pad_state.r_stick_up.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
+                                    ->GetAnalogDirectionStatus(Input::AnalogDirection::UP));
     pad_state.r_stick_down.Assign(analog_state[static_cast<std::size_t>(JoystickId::Joystick_Right)]
                                       ->GetAnalogDirectionStatus(Input::AnalogDirection::DOWN));
 
diff --git a/src/core/hle/service/nvflinger/nvflinger.cpp b/src/core/hle/service/nvflinger/nvflinger.cpp
index 134152210..437bc5dee 100644
--- a/src/core/hle/service/nvflinger/nvflinger.cpp
+++ b/src/core/hle/service/nvflinger/nvflinger.cpp
@@ -191,8 +191,6 @@ void NVFlinger::Compose() {
         // Search for a queued buffer and acquire it
         auto buffer = buffer_queue.AcquireBuffer();
 
-        MicroProfileFlip();
-
         if (!buffer) {
             continue;
         }
@@ -206,6 +204,8 @@ void NVFlinger::Compose() {
             gpu.WaitFence(fence.id, fence.value);
         }
 
+        MicroProfileFlip();
+
         // Now send the buffer to the GPU for drawing.
         // TODO(Subv): Support more than just disp0. The display device selection is probably based
         // on which display we're drawing (Default, Internal, External, etc)
diff --git a/src/core/hle/service/set/set.cpp b/src/core/hle/service/set/set.cpp
index 5bcc0b588..9e12c76fc 100644
--- a/src/core/hle/service/set/set.cpp
+++ b/src/core/hle/service/set/set.cpp
@@ -111,6 +111,14 @@ void SET::GetLanguageCode(Kernel::HLERequestContext& ctx) {
     rb.PushEnum(available_language_codes[Settings::values.language_index]);
 }
 
+void SET::GetRegionCode(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_SET, "called");
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.Push(Settings::values.region_index);
+}
+
 SET::SET() : ServiceFramework("set") {
     // clang-format off
     static const FunctionInfo functions[] = {
@@ -118,7 +126,7 @@ SET::SET() : ServiceFramework("set") {
         {1, &SET::GetAvailableLanguageCodes, "GetAvailableLanguageCodes"},
         {2, &SET::MakeLanguageCode, "MakeLanguageCode"},
         {3, &SET::GetAvailableLanguageCodeCount, "GetAvailableLanguageCodeCount"},
-        {4, nullptr, "GetRegionCode"},
+        {4, &SET::GetRegionCode, "GetRegionCode"},
         {5, &SET::GetAvailableLanguageCodes2, "GetAvailableLanguageCodes2"},
         {6, &SET::GetAvailableLanguageCodeCount2, "GetAvailableLanguageCodeCount2"},
         {7, nullptr, "GetKeyCodeMap"},
diff --git a/src/core/hle/service/set/set.h b/src/core/hle/service/set/set.h
index b154e08aa..6084b345d 100644
--- a/src/core/hle/service/set/set.h
+++ b/src/core/hle/service/set/set.h
@@ -43,6 +43,7 @@ private:
     void GetAvailableLanguageCodeCount(Kernel::HLERequestContext& ctx);
     void GetAvailableLanguageCodeCount2(Kernel::HLERequestContext& ctx);
     void GetQuestFlag(Kernel::HLERequestContext& ctx);
+    void GetRegionCode(Kernel::HLERequestContext& ctx);
 };
 
 } // namespace Service::Set
diff --git a/src/core/hle/service/sm/controller.cpp b/src/core/hle/service/sm/controller.cpp
index c45b285f8..9cca84b31 100644
--- a/src/core/hle/service/sm/controller.cpp
+++ b/src/core/hle/service/sm/controller.cpp
@@ -44,7 +44,7 @@ void Controller::QueryPointerBufferSize(Kernel::HLERequestContext& ctx) {
 
     IPC::ResponseBuilder rb{ctx, 3};
     rb.Push(RESULT_SUCCESS);
-    rb.Push<u16>(0x500);
+    rb.Push<u16>(0x1000);
 }
 
 Controller::Controller() : ServiceFramework("IpcController") {
diff --git a/src/core/hle/service/time/time_zone_content_manager.cpp b/src/core/hle/service/time/time_zone_content_manager.cpp
index 57b1a2bca..78d4acd95 100644
--- a/src/core/hle/service/time/time_zone_content_manager.cpp
+++ b/src/core/hle/service/time/time_zone_content_manager.cpp
@@ -53,7 +53,7 @@ static std::vector<std::string> BuildLocationNameCache(Core::System& system) {
         return {};
     }
 
-    std::vector<char> raw_data(binary_list->GetSize());
+    std::vector<char> raw_data(binary_list->GetSize() + 1);
     binary_list->ReadBytes<char>(raw_data.data(), binary_list->GetSize());
 
     std::stringstream data_stream{raw_data.data()};
diff --git a/src/core/loader/deconstructed_rom_directory.cpp b/src/core/loader/deconstructed_rom_directory.cpp
index d19c3623c..53559e8b1 100644
--- a/src/core/loader/deconstructed_rom_directory.cpp
+++ b/src/core/loader/deconstructed_rom_directory.cpp
@@ -129,12 +129,6 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
     }
     metadata.Print();
 
-    const FileSys::ProgramAddressSpaceType arch_bits{metadata.GetAddressSpaceType()};
-    if (arch_bits == FileSys::ProgramAddressSpaceType::Is32Bit ||
-        arch_bits == FileSys::ProgramAddressSpaceType::Is32BitNoMap) {
-        return {ResultStatus::Error32BitISA, {}};
-    }
-
     if (process.LoadFromMetadata(metadata).IsError()) {
         return {ResultStatus::ErrorUnableToParseKernelMetadata, {}};
     }
diff --git a/src/core/reporter.cpp b/src/core/reporter.cpp
index f95eee3b1..85ac81ef7 100644
--- a/src/core/reporter.cpp
+++ b/src/core/reporter.cpp
@@ -111,7 +111,7 @@ json GetProcessorStateDataAuto(Core::System& system) {
     const auto& vm_manager{process->VMManager()};
     auto& arm{system.CurrentArmInterface()};
 
-    Core::ARM_Interface::ThreadContext context{};
+    Core::ARM_Interface::ThreadContext64 context{};
     arm.SaveContext(context);
 
     return GetProcessorStateData(process->Is64BitProcess() ? "AArch64" : "AArch32",
diff --git a/src/core/settings.cpp b/src/core/settings.cpp
index d1fc94060..c1282cb80 100644
--- a/src/core/settings.cpp
+++ b/src/core/settings.cpp
@@ -86,6 +86,7 @@ void LogSettings() {
     LogSetting("System_RngSeed", Settings::values.rng_seed.value_or(0));
     LogSetting("System_CurrentUser", Settings::values.current_user);
     LogSetting("System_LanguageIndex", Settings::values.language_index);
+    LogSetting("System_RegionIndex", Settings::values.region_index);
     LogSetting("Core_UseMultiCore", Settings::values.use_multi_core);
     LogSetting("Renderer_UseResolutionFactor", Settings::values.resolution_factor);
     LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
@@ -94,6 +95,7 @@ void LogSettings() {
     LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
     LogSetting("Renderer_UseAsynchronousGpuEmulation",
                Settings::values.use_asynchronous_gpu_emulation);
+    LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
     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 f837d3fbc..79ec01731 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -387,6 +387,8 @@ struct Values {
 
     s32 current_user;
     s32 language_index;
+    s32 region_index;
+    s32 sound_index;
 
     // Controls
     std::array<PlayerInput, 10> players;
@@ -430,11 +432,13 @@ struct Values {
 
     float resolution_factor;
     int aspect_ratio;
+    int max_anisotropy;
     bool use_frame_limit;
     u16 frame_limit;
     bool use_disk_shader_cache;
     bool use_accurate_gpu_emulation;
     bool use_asynchronous_gpu_emulation;
+    bool use_vsync;
     bool force_30fps_mode;
 
     float bg_red;
diff --git a/src/core/telemetry_session.cpp b/src/core/telemetry_session.cpp
index 0e72d31cd..0f3685d1c 100644
--- a/src/core/telemetry_session.cpp
+++ b/src/core/telemetry_session.cpp
@@ -188,6 +188,7 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
              Settings::values.use_accurate_gpu_emulation);
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation);
+    AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync);
     AddField(field_type, "System_UseDockedMode", Settings::values.use_docked_mode);
 }
 
diff --git a/src/input_common/analog_from_button.cpp b/src/input_common/analog_from_button.cpp
index e1a260762..6cabdaa3c 100755
--- a/src/input_common/analog_from_button.cpp
+++ b/src/input_common/analog_from_button.cpp
@@ -34,6 +34,20 @@ public:
                                y * coef * (x == 0 ? 1.0f : SQRT_HALF));
     }
 
+    bool GetAnalogDirectionStatus(Input::AnalogDirection direction) const override {
+        switch (direction) {
+        case Input::AnalogDirection::RIGHT:
+            return right->GetStatus();
+        case Input::AnalogDirection::LEFT:
+            return left->GetStatus();
+        case Input::AnalogDirection::UP:
+            return up->GetStatus();
+        case Input::AnalogDirection::DOWN:
+            return down->GetStatus();
+        }
+        return false;
+    }
+
 private:
     Button up;
     Button down;
diff --git a/src/input_common/udp/client.cpp b/src/input_common/udp/client.cpp
index 2228571a6..da5227058 100644
--- a/src/input_common/udp/client.cpp
+++ b/src/input_common/udp/client.cpp
@@ -32,8 +32,16 @@ public:
                     SocketCallback callback)
         : callback(std::move(callback)), timer(io_service),
           socket(io_service, udp::endpoint(udp::v4(), 0)), client_id(client_id),
-          pad_index(pad_index),
-          send_endpoint(udp::endpoint(boost::asio::ip::make_address_v4(host), port)) {}
+          pad_index(pad_index) {
+        boost::system::error_code ec{};
+        auto ipv4 = boost::asio::ip::make_address_v4(host, ec);
+        if (ec.value() != boost::system::errc::success) {
+            LOG_ERROR(Input, "Invalid IPv4 address \"{}\" provided to socket", host);
+            ipv4 = boost::asio::ip::address_v4{};
+        }
+
+        send_endpoint = {udp::endpoint(ipv4, port)};
+    }
 
     void Stop() {
         io_service.stop();
@@ -85,17 +93,18 @@ private:
     }
 
     void HandleSend(const boost::system::error_code& error) {
+        boost::system::error_code _ignored{};
         // Send a request for getting port info for the pad
         Request::PortInfo port_info{1, {pad_index, 0, 0, 0}};
         const auto port_message = Request::Create(port_info, client_id);
         std::memcpy(&send_buffer1, &port_message, PORT_INFO_SIZE);
-        socket.send_to(boost::asio::buffer(send_buffer1), send_endpoint);
+        socket.send_to(boost::asio::buffer(send_buffer1), send_endpoint, {}, _ignored);
 
         // Send a request for getting pad data for the pad
         Request::PadData pad_data{Request::PadData::Flags::Id, pad_index, EMPTY_MAC_ADDRESS};
         const auto pad_message = Request::Create(pad_data, client_id);
         std::memcpy(send_buffer2.data(), &pad_message, PAD_DATA_SIZE);
-        socket.send_to(boost::asio::buffer(send_buffer2), send_endpoint);
+        socket.send_to(boost::asio::buffer(send_buffer2), send_endpoint, {}, _ignored);
         StartSend(timer.expiry());
     }
 
diff --git a/src/input_common/udp/protocol.cpp b/src/input_common/udp/protocol.cpp
index a982ac49d..5e50bd612 100644
--- a/src/input_common/udp/protocol.cpp
+++ b/src/input_common/udp/protocol.cpp
@@ -31,7 +31,6 @@ namespace Response {
  */
 std::optional<Type> Validate(u8* data, std::size_t size) {
     if (size < sizeof(Header)) {
-        LOG_DEBUG(Input, "Invalid UDP packet received");
         return std::nullopt;
     }
     Header header{};
diff --git a/src/input_common/udp/udp.cpp b/src/input_common/udp/udp.cpp
index ca99cc22f..8c6ef1394 100644
--- a/src/input_common/udp/udp.cpp
+++ b/src/input_common/udp/udp.cpp
@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <mutex>
+#include <optional>
 #include <tuple>
 
 #include "common/param_package.h"
@@ -44,7 +45,7 @@ public:
     std::unique_ptr<Input::TouchDevice> Create(const Common::ParamPackage& params) override {
         {
             std::lock_guard guard(status->update_mutex);
-            status->touch_calibration.emplace();
+            status->touch_calibration = DeviceStatus::CalibrationData{};
             // These default values work well for DS4 but probably not other touch inputs
             status->touch_calibration->min_x = params.Get("min_x", 100);
             status->touch_calibration->min_y = params.Get("min_y", 50);
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 4b0c6346f..91df062d7 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -2,6 +2,8 @@ add_library(video_core STATIC
     buffer_cache/buffer_block.h
     buffer_cache/buffer_cache.h
     buffer_cache/map_interval.h
+    dirty_flags.cpp
+    dirty_flags.h
     dma_pusher.cpp
     dma_pusher.h
     engines/const_buffer_engine_interface.h
@@ -63,14 +65,12 @@ add_library(video_core STATIC
     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
     renderer_opengl/gl_shader_manager.h
     renderer_opengl/gl_shader_util.cpp
     renderer_opengl/gl_shader_util.h
-    renderer_opengl/gl_state.cpp
-    renderer_opengl/gl_state.h
+    renderer_opengl/gl_state_tracker.cpp
+    renderer_opengl/gl_state_tracker.h
     renderer_opengl/gl_stream_buffer.cpp
     renderer_opengl/gl_stream_buffer.h
     renderer_opengl/gl_texture_cache.cpp
@@ -116,8 +116,6 @@ add_library(video_core STATIC
     shader/ast.h
     shader/compiler_settings.cpp
     shader/compiler_settings.h
-    shader/const_buffer_locker.cpp
-    shader/const_buffer_locker.h
     shader/control_flow.cpp
     shader/control_flow.h
     shader/decode.cpp
@@ -126,9 +124,13 @@ add_library(video_core STATIC
     shader/node_helper.cpp
     shader/node_helper.h
     shader/node.h
+    shader/registry.cpp
+    shader/registry.h
     shader/shader_ir.cpp
     shader/shader_ir.h
     shader/track.cpp
+    shader/transform_feedback.cpp
+    shader/transform_feedback.h
     surface.cpp
     surface.h
     texture_cache/format_lookup_table.cpp
@@ -198,6 +200,8 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_shader_util.h
         renderer_vulkan/vk_staging_buffer_pool.cpp
         renderer_vulkan/vk_staging_buffer_pool.h
+        renderer_vulkan/vk_state_tracker.cpp
+        renderer_vulkan/vk_state_tracker.h
         renderer_vulkan/vk_stream_buffer.cpp
         renderer_vulkan/vk_stream_buffer.h
         renderer_vulkan/vk_swapchain.cpp
diff --git a/src/video_core/dirty_flags.cpp b/src/video_core/dirty_flags.cpp
new file mode 100644
index 000000000..e16075993
--- /dev/null
+++ b/src/video_core/dirty_flags.cpp
@@ -0,0 +1,38 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <array>
+#include <cstddef>
+
+#include "common/common_types.h"
+#include "video_core/dirty_flags.h"
+
+#define OFF(field_name) MAXWELL3D_REG_INDEX(field_name)
+#define NUM(field_name) (sizeof(::Tegra::Engines::Maxwell3D::Regs::field_name) / sizeof(u32))
+
+namespace VideoCommon::Dirty {
+
+using Tegra::Engines::Maxwell3D;
+
+void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables) {
+    static constexpr std::size_t num_per_rt = NUM(rt[0]);
+    static constexpr std::size_t begin = OFF(rt);
+    static constexpr std::size_t num = num_per_rt * Maxwell3D::Regs::NumRenderTargets;
+    for (std::size_t rt = 0; rt < Maxwell3D::Regs::NumRenderTargets; ++rt) {
+        FillBlock(tables[0], begin + rt * num_per_rt, num_per_rt, ColorBuffer0 + rt);
+    }
+    FillBlock(tables[1], begin, num, RenderTargets);
+
+    static constexpr std::array zeta_flags{ZetaBuffer, RenderTargets};
+    for (std::size_t i = 0; i < std::size(zeta_flags); ++i) {
+        const u8 flag = zeta_flags[i];
+        auto& table = tables[i];
+        table[OFF(zeta_enable)] = flag;
+        table[OFF(zeta_width)] = flag;
+        table[OFF(zeta_height)] = flag;
+        FillBlock(table, OFF(zeta), NUM(zeta), flag);
+    }
+}
+
+} // namespace VideoCommon::Dirty
diff --git a/src/video_core/dirty_flags.h b/src/video_core/dirty_flags.h
new file mode 100644
index 000000000..3f6c1d83a
--- /dev/null
+++ b/src/video_core/dirty_flags.h
@@ -0,0 +1,49 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <cstddef>
+#include <iterator>
+
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+
+namespace VideoCommon::Dirty {
+
+enum : u8 {
+    NullEntry = 0,
+
+    RenderTargets,
+    ColorBuffer0,
+    ColorBuffer1,
+    ColorBuffer2,
+    ColorBuffer3,
+    ColorBuffer4,
+    ColorBuffer5,
+    ColorBuffer6,
+    ColorBuffer7,
+    ZetaBuffer,
+
+    LastCommonEntry,
+};
+
+template <typename Integer>
+void FillBlock(Tegra::Engines::Maxwell3D::DirtyState::Table& table, std::size_t begin,
+               std::size_t num, Integer dirty_index) {
+    const auto it = std::begin(table) + begin;
+    std::fill(it, it + num, static_cast<u8>(dirty_index));
+}
+
+template <typename Integer1, typename Integer2>
+void FillBlock(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables, std::size_t begin,
+               std::size_t num, Integer1 index_a, Integer2 index_b) {
+    FillBlock(tables[0], begin, num, index_a);
+    FillBlock(tables[1], begin, num, index_b);
+}
+
+void SetupDirtyRenderTargets(Tegra::Engines::Maxwell3D::DirtyState::Tables& tables);
+
+} // namespace VideoCommon::Dirty
diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index 0094fd715..713c14182 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -22,7 +22,7 @@ void DmaPusher::DispatchCalls() {
     MICROPROFILE_SCOPE(DispatchCalls);
 
     // On entering GPU code, assume all memory may be touched by the ARM core.
-    gpu.Maxwell3D().dirty.OnMemoryWrite();
+    gpu.Maxwell3D().OnMemoryWrite();
 
     dma_pushbuffer_subindex = 0;
 
diff --git a/src/video_core/engines/const_buffer_engine_interface.h b/src/video_core/engines/const_buffer_engine_interface.h
index d56a47710..724ee0fd6 100644
--- a/src/video_core/engines/const_buffer_engine_interface.h
+++ b/src/video_core/engines/const_buffer_engine_interface.h
@@ -16,11 +16,12 @@ namespace Tegra::Engines {
 
 struct SamplerDescriptor {
     union {
-        BitField<0, 20, Tegra::Shader::TextureType> texture_type;
-        BitField<20, 1, u32> is_array;
-        BitField<21, 1, u32> is_buffer;
-        BitField<22, 1, u32> is_shadow;
-        u32 raw{};
+        u32 raw = 0;
+        BitField<0, 2, Tegra::Shader::TextureType> texture_type;
+        BitField<2, 3, Tegra::Texture::ComponentType> component_type;
+        BitField<5, 1, u32> is_array;
+        BitField<6, 1, u32> is_buffer;
+        BitField<7, 1, u32> is_shadow;
     };
 
     bool operator==(const SamplerDescriptor& rhs) const noexcept {
@@ -31,68 +32,48 @@ struct SamplerDescriptor {
         return !operator==(rhs);
     }
 
-    static SamplerDescriptor FromTicTexture(Tegra::Texture::TextureType tic_texture_type) {
+    static SamplerDescriptor FromTIC(const Tegra::Texture::TICEntry& tic) {
+        using Tegra::Shader::TextureType;
         SamplerDescriptor result;
-        switch (tic_texture_type) {
+
+        // This is going to be used to determine the shading language type.
+        // Because of that we don't care about all component types on color textures.
+        result.component_type.Assign(tic.r_type.Value());
+
+        switch (tic.texture_type.Value()) {
         case Tegra::Texture::TextureType::Texture1D:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture1D);
-            result.is_array.Assign(0);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
+            result.texture_type.Assign(TextureType::Texture1D);
             return result;
         case Tegra::Texture::TextureType::Texture2D:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
-            result.is_array.Assign(0);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
+            result.texture_type.Assign(TextureType::Texture2D);
             return result;
         case Tegra::Texture::TextureType::Texture3D:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture3D);
-            result.is_array.Assign(0);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
+            result.texture_type.Assign(TextureType::Texture3D);
             return result;
         case Tegra::Texture::TextureType::TextureCubemap:
-            result.texture_type.Assign(Tegra::Shader::TextureType::TextureCube);
-            result.is_array.Assign(0);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
+            result.texture_type.Assign(TextureType::TextureCube);
             return result;
         case Tegra::Texture::TextureType::Texture1DArray:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture1D);
+            result.texture_type.Assign(TextureType::Texture1D);
             result.is_array.Assign(1);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
             return result;
         case Tegra::Texture::TextureType::Texture2DArray:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
+            result.texture_type.Assign(TextureType::Texture2D);
             result.is_array.Assign(1);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
             return result;
         case Tegra::Texture::TextureType::Texture1DBuffer:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture1D);
-            result.is_array.Assign(0);
+            result.texture_type.Assign(TextureType::Texture1D);
             result.is_buffer.Assign(1);
-            result.is_shadow.Assign(0);
             return result;
         case Tegra::Texture::TextureType::Texture2DNoMipmap:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
-            result.is_array.Assign(0);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
+            result.texture_type.Assign(TextureType::Texture2D);
             return result;
         case Tegra::Texture::TextureType::TextureCubeArray:
-            result.texture_type.Assign(Tegra::Shader::TextureType::TextureCube);
+            result.texture_type.Assign(TextureType::TextureCube);
             result.is_array.Assign(1);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
             return result;
         default:
-            result.texture_type.Assign(Tegra::Shader::TextureType::Texture2D);
-            result.is_array.Assign(0);
-            result.is_buffer.Assign(0);
-            result.is_shadow.Assign(0);
+            result.texture_type.Assign(TextureType::Texture2D);
             return result;
         }
     }
diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp
index 4b824aa4e..368c75a66 100644
--- a/src/video_core/engines/kepler_compute.cpp
+++ b/src/video_core/engines/kepler_compute.cpp
@@ -39,7 +39,7 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
         const bool is_last_call = method_call.IsLastCall();
         upload_state.ProcessData(method_call.argument, is_last_call);
         if (is_last_call) {
-            system.GPU().Maxwell3D().dirty.OnMemoryWrite();
+            system.GPU().Maxwell3D().OnMemoryWrite();
         }
         break;
     }
@@ -89,7 +89,7 @@ SamplerDescriptor KeplerCompute::AccessBindlessSampler(ShaderType stage, u64 con
 
     const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
     const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle);
-    SamplerDescriptor result = SamplerDescriptor::FromTicTexture(tex_info.tic.texture_type.Value());
+    SamplerDescriptor result = SamplerDescriptor::FromTIC(tex_info.tic);
     result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value());
     return result;
 }
@@ -119,14 +119,6 @@ Texture::TICEntry KeplerCompute::GetTICEntry(u32 tic_index) const {
     Texture::TICEntry tic_entry;
     memory_manager.ReadBlockUnsafe(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry));
 
-    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
-    DEBUG_ASSERT(r_type == g_type && r_type == b_type && r_type == a_type);
-
     return tic_entry;
 }
 
diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp
index fa4a7c5c1..597872e43 100644
--- a/src/video_core/engines/kepler_memory.cpp
+++ b/src/video_core/engines/kepler_memory.cpp
@@ -34,7 +34,7 @@ void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) {
         const bool is_last_call = method_call.IsLastCall();
         upload_state.ProcessData(method_call.argument, is_last_call);
         if (is_last_call) {
-            system.GPU().Maxwell3D().dirty.OnMemoryWrite();
+            system.GPU().Maxwell3D().OnMemoryWrite();
         }
         break;
     }
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index b28de1092..ba63b44b4 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -26,7 +26,8 @@ Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& raste
                      MemoryManager& memory_manager)
     : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager},
       macro_interpreter{*this}, upload_state{memory_manager, regs.upload} {
-    InitDirtySettings();
+    dirty.flags.flip();
+
     InitializeRegisterDefaults();
 }
 
@@ -75,8 +76,8 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.stencil_back_mask = 0xFFFFFFFF;
 
     regs.depth_test_func = Regs::ComparisonOp::Always;
-    regs.cull.front_face = Regs::Cull::FrontFace::CounterClockWise;
-    regs.cull.cull_face = Regs::Cull::CullFace::Back;
+    regs.front_face = Regs::FrontFace::CounterClockWise;
+    regs.cull_face = Regs::CullFace::Back;
 
     // TODO(Rodrigo): Most games do not set a point size. I think this is a case of a
     // register carrying a default value. Assume it's OpenGL's default (1).
@@ -95,7 +96,9 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.rasterize_enable = 1;
     regs.rt_separate_frag_data = 1;
     regs.framebuffer_srgb = 1;
-    regs.cull.front_face = Maxwell3D::Regs::Cull::FrontFace::ClockWise;
+    regs.front_face = Maxwell3D::Regs::FrontFace::ClockWise;
+
+    shadow_state = regs;
 
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_end_gl)] = true;
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_begin_gl)] = true;
@@ -103,164 +106,6 @@ void Maxwell3D::InitializeRegisterDefaults() {
     mme_inline[MAXWELL3D_REG_INDEX(index_array.count)] = true;
 }
 
-#define DIRTY_REGS_POS(field_name) static_cast<u8>(offsetof(Maxwell3D::DirtyRegs, field_name))
-
-void Maxwell3D::InitDirtySettings() {
-    const auto set_block = [this](std::size_t start, std::size_t range, u8 position) {
-        const auto start_itr = dirty_pointers.begin() + start;
-        const auto end_itr = start_itr + range;
-        std::fill(start_itr, end_itr, position);
-    };
-    dirty.regs.fill(true);
-
-    // Init Render Targets
-    constexpr u32 registers_per_rt = sizeof(regs.rt[0]) / sizeof(u32);
-    constexpr u32 rt_start_reg = MAXWELL3D_REG_INDEX(rt);
-    constexpr u32 rt_end_reg = rt_start_reg + registers_per_rt * 8;
-    u8 rt_dirty_reg = DIRTY_REGS_POS(render_target);
-    for (u32 rt_reg = rt_start_reg; rt_reg < rt_end_reg; rt_reg += registers_per_rt) {
-        set_block(rt_reg, registers_per_rt, rt_dirty_reg);
-        ++rt_dirty_reg;
-    }
-    constexpr u32 depth_buffer_flag = DIRTY_REGS_POS(depth_buffer);
-    dirty_pointers[MAXWELL3D_REG_INDEX(zeta_enable)] = depth_buffer_flag;
-    dirty_pointers[MAXWELL3D_REG_INDEX(zeta_width)] = depth_buffer_flag;
-    dirty_pointers[MAXWELL3D_REG_INDEX(zeta_height)] = depth_buffer_flag;
-    constexpr u32 registers_in_zeta = sizeof(regs.zeta) / sizeof(u32);
-    constexpr u32 zeta_reg = MAXWELL3D_REG_INDEX(zeta);
-    set_block(zeta_reg, registers_in_zeta, depth_buffer_flag);
-
-    // Init Vertex Arrays
-    constexpr u32 vertex_array_start = MAXWELL3D_REG_INDEX(vertex_array);
-    constexpr u32 vertex_array_size = sizeof(regs.vertex_array[0]) / sizeof(u32);
-    constexpr u32 vertex_array_end = vertex_array_start + vertex_array_size * Regs::NumVertexArrays;
-    u8 va_dirty_reg = DIRTY_REGS_POS(vertex_array);
-    u8 vi_dirty_reg = DIRTY_REGS_POS(vertex_instance);
-    for (u32 vertex_reg = vertex_array_start; vertex_reg < vertex_array_end;
-         vertex_reg += vertex_array_size) {
-        set_block(vertex_reg, 3, va_dirty_reg);
-        // The divisor concerns vertex array instances
-        dirty_pointers[static_cast<std::size_t>(vertex_reg) + 3] = vi_dirty_reg;
-        ++va_dirty_reg;
-        ++vi_dirty_reg;
-    }
-    constexpr u32 vertex_limit_start = MAXWELL3D_REG_INDEX(vertex_array_limit);
-    constexpr u32 vertex_limit_size = sizeof(regs.vertex_array_limit[0]) / sizeof(u32);
-    constexpr u32 vertex_limit_end = vertex_limit_start + vertex_limit_size * Regs::NumVertexArrays;
-    va_dirty_reg = DIRTY_REGS_POS(vertex_array);
-    for (u32 vertex_reg = vertex_limit_start; vertex_reg < vertex_limit_end;
-         vertex_reg += vertex_limit_size) {
-        set_block(vertex_reg, vertex_limit_size, va_dirty_reg);
-        va_dirty_reg++;
-    }
-    constexpr u32 vertex_instance_start = MAXWELL3D_REG_INDEX(instanced_arrays);
-    constexpr u32 vertex_instance_size =
-        sizeof(regs.instanced_arrays.is_instanced[0]) / sizeof(u32);
-    constexpr u32 vertex_instance_end =
-        vertex_instance_start + vertex_instance_size * Regs::NumVertexArrays;
-    vi_dirty_reg = DIRTY_REGS_POS(vertex_instance);
-    for (u32 vertex_reg = vertex_instance_start; vertex_reg < vertex_instance_end;
-         vertex_reg += vertex_instance_size) {
-        set_block(vertex_reg, vertex_instance_size, vi_dirty_reg);
-        vi_dirty_reg++;
-    }
-    set_block(MAXWELL3D_REG_INDEX(vertex_attrib_format), regs.vertex_attrib_format.size(),
-              DIRTY_REGS_POS(vertex_attrib_format));
-
-    // Init Shaders
-    constexpr u32 shader_registers_count =
-        sizeof(regs.shader_config[0]) * Regs::MaxShaderProgram / sizeof(u32);
-    set_block(MAXWELL3D_REG_INDEX(shader_config[0]), shader_registers_count,
-              DIRTY_REGS_POS(shaders));
-
-    // State
-
-    // Viewport
-    constexpr u8 viewport_dirty_reg = DIRTY_REGS_POS(viewport);
-    constexpr u32 viewport_start = MAXWELL3D_REG_INDEX(viewports);
-    constexpr u32 viewport_size = sizeof(regs.viewports) / sizeof(u32);
-    set_block(viewport_start, viewport_size, viewport_dirty_reg);
-    constexpr u32 view_volume_start = MAXWELL3D_REG_INDEX(view_volume_clip_control);
-    constexpr u32 view_volume_size = sizeof(regs.view_volume_clip_control) / sizeof(u32);
-    set_block(view_volume_start, view_volume_size, viewport_dirty_reg);
-
-    // Viewport transformation
-    constexpr u32 viewport_trans_start = MAXWELL3D_REG_INDEX(viewport_transform);
-    constexpr u32 viewport_trans_size = sizeof(regs.viewport_transform) / sizeof(u32);
-    set_block(viewport_trans_start, viewport_trans_size, DIRTY_REGS_POS(viewport_transform));
-
-    // Cullmode
-    constexpr u32 cull_mode_start = MAXWELL3D_REG_INDEX(cull);
-    constexpr u32 cull_mode_size = sizeof(regs.cull) / sizeof(u32);
-    set_block(cull_mode_start, cull_mode_size, DIRTY_REGS_POS(cull_mode));
-
-    // Screen y control
-    dirty_pointers[MAXWELL3D_REG_INDEX(screen_y_control)] = DIRTY_REGS_POS(screen_y_control);
-
-    // Primitive Restart
-    constexpr u32 primitive_restart_start = MAXWELL3D_REG_INDEX(primitive_restart);
-    constexpr u32 primitive_restart_size = sizeof(regs.primitive_restart) / sizeof(u32);
-    set_block(primitive_restart_start, primitive_restart_size, DIRTY_REGS_POS(primitive_restart));
-
-    // Depth Test
-    constexpr u8 depth_test_dirty_reg = DIRTY_REGS_POS(depth_test);
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_test_enable)] = depth_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_write_enabled)] = depth_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_test_func)] = depth_test_dirty_reg;
-
-    // Stencil Test
-    constexpr u32 stencil_test_dirty_reg = DIRTY_REGS_POS(stencil_test);
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_enable)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_func_func)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_func_ref)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_func_mask)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_op_fail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_op_zfail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_op_zpass)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_front_mask)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_two_side_enable)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_func_func)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_func_ref)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_func_mask)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_op_fail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_op_zfail)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_op_zpass)] = stencil_test_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(stencil_back_mask)] = stencil_test_dirty_reg;
-
-    // Color Mask
-    constexpr u8 color_mask_dirty_reg = DIRTY_REGS_POS(color_mask);
-    dirty_pointers[MAXWELL3D_REG_INDEX(color_mask_common)] = color_mask_dirty_reg;
-    set_block(MAXWELL3D_REG_INDEX(color_mask), sizeof(regs.color_mask) / sizeof(u32),
-              color_mask_dirty_reg);
-    // Blend State
-    constexpr u8 blend_state_dirty_reg = DIRTY_REGS_POS(blend_state);
-    set_block(MAXWELL3D_REG_INDEX(blend_color), sizeof(regs.blend_color) / sizeof(u32),
-              blend_state_dirty_reg);
-    dirty_pointers[MAXWELL3D_REG_INDEX(independent_blend_enable)] = blend_state_dirty_reg;
-    set_block(MAXWELL3D_REG_INDEX(blend), sizeof(regs.blend) / sizeof(u32), blend_state_dirty_reg);
-    set_block(MAXWELL3D_REG_INDEX(independent_blend), sizeof(regs.independent_blend) / sizeof(u32),
-              blend_state_dirty_reg);
-
-    // Scissor State
-    constexpr u8 scissor_test_dirty_reg = DIRTY_REGS_POS(scissor_test);
-    set_block(MAXWELL3D_REG_INDEX(scissor_test), sizeof(regs.scissor_test) / sizeof(u32),
-              scissor_test_dirty_reg);
-
-    // Polygon Offset
-    constexpr u8 polygon_offset_dirty_reg = DIRTY_REGS_POS(polygon_offset);
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_fill_enable)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_line_enable)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_point_enable)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_units)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_factor)] = polygon_offset_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_clamp)] = polygon_offset_dirty_reg;
-
-    // Depth bounds
-    constexpr u8 depth_bounds_values_dirty_reg = DIRTY_REGS_POS(depth_bounds_values);
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_bounds[0])] = depth_bounds_values_dirty_reg;
-    dirty_pointers[MAXWELL3D_REG_INDEX(depth_bounds[1])] = depth_bounds_values_dirty_reg;
-}
-
 void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters) {
     // Reset the current macro.
     executing_macro = 0;
@@ -317,31 +162,34 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid Maxwell3D register, increase the size of the Regs structure");
 
-    if (regs.reg_array[method] != method_call.argument) {
-        regs.reg_array[method] = method_call.argument;
-        const std::size_t dirty_reg = dirty_pointers[method];
-        if (dirty_reg) {
-            dirty.regs[dirty_reg] = true;
-            if (dirty_reg >= DIRTY_REGS_POS(vertex_array) &&
-                dirty_reg < DIRTY_REGS_POS(vertex_array_buffers)) {
-                dirty.vertex_array_buffers = true;
-            } else if (dirty_reg >= DIRTY_REGS_POS(vertex_instance) &&
-                       dirty_reg < DIRTY_REGS_POS(vertex_instances)) {
-                dirty.vertex_instances = true;
-            } else if (dirty_reg >= DIRTY_REGS_POS(render_target) &&
-                       dirty_reg < DIRTY_REGS_POS(render_settings)) {
-                dirty.render_settings = true;
-            }
+    u32 arg = method_call.argument;
+    // Keep track of the register value in shadow_state when requested.
+    if (shadow_state.shadow_ram_control == Regs::ShadowRamControl::Track ||
+        shadow_state.shadow_ram_control == Regs::ShadowRamControl::TrackWithFilter) {
+        shadow_state.reg_array[method] = arg;
+    } else if (shadow_state.shadow_ram_control == Regs::ShadowRamControl::Replay) {
+        arg = shadow_state.reg_array[method];
+    }
+
+    if (regs.reg_array[method] != arg) {
+        regs.reg_array[method] = arg;
+
+        for (const auto& table : dirty.tables) {
+            dirty.flags[table[method]] = true;
         }
     }
 
     switch (method) {
+    case MAXWELL3D_REG_INDEX(shadow_ram_control): {
+        shadow_state.shadow_ram_control = static_cast<Regs::ShadowRamControl>(method_call.argument);
+        break;
+    }
     case MAXWELL3D_REG_INDEX(macros.data): {
-        ProcessMacroUpload(method_call.argument);
+        ProcessMacroUpload(arg);
         break;
     }
     case MAXWELL3D_REG_INDEX(macros.bind): {
-        ProcessMacroBind(method_call.argument);
+        ProcessMacroBind(arg);
         break;
     }
     case MAXWELL3D_REG_INDEX(firmware[4]): {
@@ -417,9 +265,9 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     }
     case MAXWELL3D_REG_INDEX(data_upload): {
         const bool is_last_call = method_call.IsLastCall();
-        upload_state.ProcessData(method_call.argument, is_last_call);
+        upload_state.ProcessData(arg, is_last_call);
         if (is_last_call) {
-            dirty.OnMemoryWrite();
+            OnMemoryWrite();
         }
         break;
     }
@@ -727,7 +575,7 @@ void Maxwell3D::FinishCBData() {
 
     const u32 id = cb_data_state.id;
     memory_manager.WriteBlock(address, cb_data_state.buffer[id].data(), size);
-    dirty.OnMemoryWrite();
+    OnMemoryWrite();
 
     cb_data_state.id = null_cb_data;
     cb_data_state.current = null_cb_data;
@@ -805,7 +653,7 @@ SamplerDescriptor Maxwell3D::AccessBindlessSampler(ShaderType stage, u64 const_b
 
     const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(tex_info_address)};
     const Texture::FullTextureInfo tex_info = GetTextureInfo(tex_handle);
-    SamplerDescriptor result = SamplerDescriptor::FromTicTexture(tex_info.tic.texture_type.Value());
+    SamplerDescriptor result = SamplerDescriptor::FromTIC(tex_info.tic);
     result.is_shadow.Assign(tex_info.tsc.depth_compare_enabled.Value());
     return result;
 }
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 26939be3f..d24c9f657 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -6,6 +6,7 @@
 
 #include <array>
 #include <bitset>
+#include <limits>
 #include <optional>
 #include <type_traits>
 #include <unordered_map>
@@ -66,6 +67,7 @@ public:
         static constexpr std::size_t NumVaryings = 31;
         static constexpr std::size_t NumImages = 8; // TODO(Rodrigo): Investigate this number
         static constexpr std::size_t NumClipDistances = 8;
+        static constexpr std::size_t NumTransformFeedbackBuffers = 4;
         static constexpr std::size_t MaxShaderProgram = 6;
         static constexpr std::size_t MaxShaderStage = 5;
         // Maximum number of const buffers per shader stage.
@@ -431,21 +433,15 @@ public:
             GeneratedPrimitives = 0x1F,
         };
 
-        struct Cull {
-            enum class FrontFace : u32 {
-                ClockWise = 0x0900,
-                CounterClockWise = 0x0901,
-            };
-
-            enum class CullFace : u32 {
-                Front = 0x0404,
-                Back = 0x0405,
-                FrontAndBack = 0x0408,
-            };
+        enum class FrontFace : u32 {
+            ClockWise = 0x0900,
+            CounterClockWise = 0x0901,
+        };
 
-            u32 enabled;
-            FrontFace front_face;
-            CullFace cull_face;
+        enum class CullFace : u32 {
+            Front = 0x0404,
+            Back = 0x0405,
+            FrontAndBack = 0x0408,
         };
 
         struct Blend {
@@ -529,6 +525,23 @@ public:
             FractionalEven = 2,
         };
 
+        enum class PolygonMode : u32 {
+            Point = 0x1b00,
+            Line = 0x1b01,
+            Fill = 0x1b02,
+        };
+
+        enum class ShadowRamControl : u32 {
+            // write value to shadow ram
+            Track = 0,
+            // write value to shadow ram ( with validation ??? )
+            TrackWithFilter = 1,
+            // only write to real hw register
+            Passthrough = 2,
+            // write value from shadow ram to real hw register
+            Replay = 3,
+        };
+
         struct RenderTargetConfig {
             u32 address_high;
             u32 address_low;
@@ -542,7 +555,7 @@ public:
                 BitField<12, 1, InvMemoryLayout> type;
             } memory_layout;
             union {
-                BitField<0, 16, u32> array_mode;
+                BitField<0, 16, u32> layers;
                 BitField<16, 1, u32> volume;
             };
             u32 layer_stride;
@@ -574,7 +587,7 @@ public:
             f32 translate_z;
             INSERT_UNION_PADDING_WORDS(2);
 
-            Common::Rectangle<s32> GetRect() const {
+            Common::Rectangle<f32> GetRect() const {
                 return {
                     GetX(),               // left
                     GetY() + GetHeight(), // top
@@ -583,20 +596,20 @@ public:
                 };
             };
 
-            s32 GetX() const {
-                return static_cast<s32>(std::max(0.0f, translate_x - std::fabs(scale_x)));
+            f32 GetX() const {
+                return std::max(0.0f, translate_x - std::fabs(scale_x));
             }
 
-            s32 GetY() const {
-                return static_cast<s32>(std::max(0.0f, translate_y - std::fabs(scale_y)));
+            f32 GetY() const {
+                return std::max(0.0f, translate_y - std::fabs(scale_y));
             }
 
-            s32 GetWidth() const {
-                return static_cast<s32>(translate_x + std::fabs(scale_x)) - GetX();
+            f32 GetWidth() const {
+                return translate_x + std::fabs(scale_x) - GetX();
             }
 
-            s32 GetHeight() const {
-                return static_cast<s32>(translate_y + std::fabs(scale_y)) - GetY();
+            f32 GetHeight() const {
+                return translate_y + std::fabs(scale_y) - GetY();
             }
         };
 
@@ -626,6 +639,29 @@ public:
             float depth_range_far;
         };
 
+        struct TransformFeedbackBinding {
+            u32 buffer_enable;
+            u32 address_high;
+            u32 address_low;
+            s32 buffer_size;
+            s32 buffer_offset;
+            INSERT_UNION_PADDING_WORDS(3);
+
+            GPUVAddr Address() const {
+                return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+                                             address_low);
+            }
+        };
+        static_assert(sizeof(TransformFeedbackBinding) == 32);
+
+        struct TransformFeedbackLayout {
+            u32 stream;
+            u32 varying_count;
+            u32 stride;
+            INSERT_UNION_PADDING_WORDS(1);
+        };
+        static_assert(sizeof(TransformFeedbackLayout) == 16);
+
         bool IsShaderConfigEnabled(std::size_t index) const {
             // The VertexB is always enabled.
             if (index == static_cast<std::size_t>(Regs::ShaderProgram::VertexB)) {
@@ -634,6 +670,10 @@ public:
             return shader_config[index].enable != 0;
         }
 
+        bool IsShaderConfigEnabled(Regs::ShaderProgram type) const {
+            return IsShaderConfigEnabled(static_cast<std::size_t>(type));
+        }
+
         union {
             struct {
                 INSERT_UNION_PADDING_WORDS(0x45);
@@ -645,7 +685,9 @@ public:
                     u32 bind;
                 } macros;
 
-                INSERT_UNION_PADDING_WORDS(0x17);
+                ShadowRamControl shadow_ram_control;
+
+                INSERT_UNION_PADDING_WORDS(0x16);
 
                 Upload::Registers upload;
                 struct {
@@ -682,7 +724,13 @@ public:
 
                 u32 rasterize_enable;
 
-                INSERT_UNION_PADDING_WORDS(0xF1);
+                std::array<TransformFeedbackBinding, NumTransformFeedbackBuffers> tfb_bindings;
+
+                INSERT_UNION_PADDING_WORDS(0xC0);
+
+                std::array<TransformFeedbackLayout, NumTransformFeedbackBuffers> tfb_layouts;
+
+                INSERT_UNION_PADDING_WORDS(0x1);
 
                 u32 tfb_enabled;
 
@@ -710,7 +758,12 @@ public:
 
                 s32 clear_stencil;
 
-                INSERT_UNION_PADDING_WORDS(0x7);
+                INSERT_UNION_PADDING_WORDS(0x2);
+
+                PolygonMode polygon_mode_front;
+                PolygonMode polygon_mode_back;
+
+                INSERT_UNION_PADDING_WORDS(0x3);
 
                 u32 polygon_offset_point_enable;
                 u32 polygon_offset_line_enable;
@@ -769,7 +822,11 @@ public:
                     BitField<12, 4, u32> viewport;
                 } clear_flags;
 
-                INSERT_UNION_PADDING_WORDS(0x19);
+                INSERT_UNION_PADDING_WORDS(0x10);
+
+                u32 fill_rectangle;
+
+                INSERT_UNION_PADDING_WORDS(0x8);
 
                 std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format;
 
@@ -800,8 +857,12 @@ public:
 
                 u32 zeta_width;
                 u32 zeta_height;
+                union {
+                    BitField<0, 16, u32> zeta_layers;
+                    BitField<16, 1, u32> zeta_volume;
+                };
 
-                INSERT_UNION_PADDING_WORDS(0x27);
+                INSERT_UNION_PADDING_WORDS(0x26);
 
                 u32 depth_test_enable;
 
@@ -868,16 +929,7 @@ public:
 
                 INSERT_UNION_PADDING_WORDS(0x35);
 
-                union {
-                    BitField<0, 1, u32> c0;
-                    BitField<1, 1, u32> c1;
-                    BitField<2, 1, u32> c2;
-                    BitField<3, 1, u32> c3;
-                    BitField<4, 1, u32> c4;
-                    BitField<5, 1, u32> c5;
-                    BitField<6, 1, u32> c6;
-                    BitField<7, 1, u32> c7;
-                } clip_distance_enabled;
+                u32 clip_distance_enabled;
 
                 u32 samplecnt_enable;
 
@@ -1056,7 +1108,9 @@ public:
 
                 INSERT_UNION_PADDING_WORDS(1);
 
-                Cull cull;
+                u32 cull_test_enabled;
+                FrontFace front_face;
+                CullFace cull_face;
 
                 u32 pixel_center_integer;
 
@@ -1195,7 +1249,11 @@ public:
 
                 u32 tex_cb_index;
 
-                INSERT_UNION_PADDING_WORDS(0x395);
+                INSERT_UNION_PADDING_WORDS(0x7D);
+
+                std::array<std::array<u8, 128>, NumTransformFeedbackBuffers> tfb_varying_locs;
+
+                INSERT_UNION_PADDING_WORDS(0x298);
 
                 struct {
                     /// Compressed address of a buffer that holds information about bound SSBOs.
@@ -1218,7 +1276,12 @@ public:
             };
             std::array<u32, NUM_REGS> reg_array;
         };
-    } regs{};
+    };
+
+    Regs regs{};
+
+    /// Store temporary hw register values, used by some calls to restore state after a operation
+    Regs shadow_state;
 
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32), "Maxwell3D Regs has wrong size");
     static_assert(std::is_trivially_copyable_v<Regs>, "Maxwell3D Regs must be trivially copyable");
@@ -1234,79 +1297,6 @@ public:
 
     State state{};
 
-    struct DirtyRegs {
-        static constexpr std::size_t NUM_REGS = 256;
-        static_assert(NUM_REGS - 1 <= std::numeric_limits<u8>::max());
-
-        union {
-            struct {
-                bool null_dirty;
-
-                // Vertex Attributes
-                bool vertex_attrib_format;
-
-                // Vertex Arrays
-                std::array<bool, 32> vertex_array;
-
-                bool vertex_array_buffers;
-
-                // Vertex Instances
-                std::array<bool, 32> vertex_instance;
-
-                bool vertex_instances;
-
-                // Render Targets
-                std::array<bool, 8> render_target;
-                bool depth_buffer;
-
-                bool render_settings;
-
-                // Shaders
-                bool shaders;
-
-                // Rasterizer State
-                bool viewport;
-                bool clip_coefficient;
-                bool cull_mode;
-                bool primitive_restart;
-                bool depth_test;
-                bool stencil_test;
-                bool blend_state;
-                bool scissor_test;
-                bool transform_feedback;
-                bool color_mask;
-                bool polygon_offset;
-                bool depth_bounds_values;
-
-                // Complementary
-                bool viewport_transform;
-                bool screen_y_control;
-
-                bool memory_general;
-            };
-            std::array<bool, NUM_REGS> regs;
-        };
-
-        void ResetVertexArrays() {
-            vertex_array.fill(true);
-            vertex_array_buffers = true;
-        }
-
-        void ResetRenderTargets() {
-            depth_buffer = true;
-            render_target.fill(true);
-            render_settings = true;
-        }
-
-        void OnMemoryWrite() {
-            shaders = true;
-            memory_general = true;
-            ResetRenderTargets();
-            ResetVertexArrays();
-        }
-
-    } dirty{};
-
     /// Reads a register value located at the input method address
     u32 GetRegisterValue(u32 method) const;
 
@@ -1352,6 +1342,11 @@ public:
         return execute_on;
     }
 
+    /// Notify a memory write has happened.
+    void OnMemoryWrite() {
+        dirty.flags |= dirty.on_write_stores;
+    }
+
     enum class MMEDrawMode : u32 {
         Undefined,
         Array,
@@ -1367,6 +1362,16 @@ public:
         u32 gl_end_count{};
     } mme_draw;
 
+    struct DirtyState {
+        using Flags = std::bitset<std::numeric_limits<u8>::max()>;
+        using Table = std::array<u8, Regs::NUM_REGS>;
+        using Tables = std::array<Table, 2>;
+
+        Flags flags;
+        Flags on_write_stores;
+        Tables tables{};
+    } dirty;
+
 private:
     void InitializeRegisterDefaults();
 
@@ -1413,8 +1418,6 @@ private:
     /// Retrieves information about a specific TSC entry from the TSC buffer.
     Texture::TSCEntry GetTSCEntry(u32 tsc_index) const;
 
-    void InitDirtySettings();
-
     /**
      * Call a macro on this engine.
      * @param method Method to call
@@ -1473,6 +1476,7 @@ private:
                   "Field " #field_name " has invalid position")
 
 ASSERT_REG_POSITION(macros, 0x45);
+ASSERT_REG_POSITION(shadow_ram_control, 0x49);
 ASSERT_REG_POSITION(upload, 0x60);
 ASSERT_REG_POSITION(exec_upload, 0x6C);
 ASSERT_REG_POSITION(data_upload, 0x6D);
@@ -1481,6 +1485,8 @@ ASSERT_REG_POSITION(tess_mode, 0xC8);
 ASSERT_REG_POSITION(tess_level_outer, 0xC9);
 ASSERT_REG_POSITION(tess_level_inner, 0xCD);
 ASSERT_REG_POSITION(rasterize_enable, 0xDF);
+ASSERT_REG_POSITION(tfb_bindings, 0xE0);
+ASSERT_REG_POSITION(tfb_layouts, 0x1C0);
 ASSERT_REG_POSITION(tfb_enabled, 0x1D1);
 ASSERT_REG_POSITION(rt, 0x200);
 ASSERT_REG_POSITION(viewport_transform, 0x280);
@@ -1490,6 +1496,8 @@ ASSERT_REG_POSITION(depth_mode, 0x35F);
 ASSERT_REG_POSITION(clear_color[0], 0x360);
 ASSERT_REG_POSITION(clear_depth, 0x364);
 ASSERT_REG_POSITION(clear_stencil, 0x368);
+ASSERT_REG_POSITION(polygon_mode_front, 0x36B);
+ASSERT_REG_POSITION(polygon_mode_back, 0x36C);
 ASSERT_REG_POSITION(polygon_offset_point_enable, 0x370);
 ASSERT_REG_POSITION(polygon_offset_line_enable, 0x371);
 ASSERT_REG_POSITION(polygon_offset_fill_enable, 0x372);
@@ -1503,10 +1511,12 @@ ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
 ASSERT_REG_POSITION(depth_bounds, 0x3E7);
 ASSERT_REG_POSITION(zeta, 0x3F8);
 ASSERT_REG_POSITION(clear_flags, 0x43E);
+ASSERT_REG_POSITION(fill_rectangle, 0x44F);
 ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
 ASSERT_REG_POSITION(rt_control, 0x487);
 ASSERT_REG_POSITION(zeta_width, 0x48a);
 ASSERT_REG_POSITION(zeta_height, 0x48b);
+ASSERT_REG_POSITION(zeta_layers, 0x48c);
 ASSERT_REG_POSITION(depth_test_enable, 0x4B3);
 ASSERT_REG_POSITION(independent_blend_enable, 0x4B9);
 ASSERT_REG_POSITION(depth_write_enabled, 0x4BA);
@@ -1556,7 +1566,9 @@ ASSERT_REG_POSITION(index_array, 0x5F2);
 ASSERT_REG_POSITION(polygon_offset_clamp, 0x61F);
 ASSERT_REG_POSITION(instanced_arrays, 0x620);
 ASSERT_REG_POSITION(vp_point_size, 0x644);
-ASSERT_REG_POSITION(cull, 0x646);
+ASSERT_REG_POSITION(cull_test_enabled, 0x646);
+ASSERT_REG_POSITION(front_face, 0x647);
+ASSERT_REG_POSITION(cull_face, 0x648);
 ASSERT_REG_POSITION(pixel_center_integer, 0x649);
 ASSERT_REG_POSITION(viewport_transform_enabled, 0x64B);
 ASSERT_REG_POSITION(view_volume_clip_control, 0x64F);
@@ -1573,6 +1585,7 @@ ASSERT_REG_POSITION(firmware, 0x8C0);
 ASSERT_REG_POSITION(const_buffer, 0x8E0);
 ASSERT_REG_POSITION(cb_bind[0], 0x904);
 ASSERT_REG_POSITION(tex_cb_index, 0x982);
+ASSERT_REG_POSITION(tfb_varying_locs, 0xA00);
 ASSERT_REG_POSITION(ssbo_info, 0xD18);
 ASSERT_REG_POSITION(tex_info_buffers.address[0], 0xD2A);
 ASSERT_REG_POSITION(tex_info_buffers.size[0], 0xD2F);
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp
index ad8453c5f..c2610f992 100644
--- a/src/video_core/engines/maxwell_dma.cpp
+++ b/src/video_core/engines/maxwell_dma.cpp
@@ -57,7 +57,7 @@ void MaxwellDMA::HandleCopy() {
     }
 
     // All copies here update the main memory, so mark all rasterizer states as invalid.
-    system.GPU().Maxwell3D().dirty.OnMemoryWrite();
+    system.GPU().Maxwell3D().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
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index c9bc83cd7..49dc5abe0 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -82,6 +82,10 @@ union Attribute {
         Position = 7,
         Attribute_0 = 8,
         Attribute_31 = 39,
+        FrontColor = 40,
+        FrontSecondaryColor = 41,
+        BackColor = 42,
+        BackSecondaryColor = 43,
         ClipDistances0123 = 44,
         ClipDistances4567 = 45,
         PointCoord = 46,
@@ -89,6 +93,8 @@ union Attribute {
         // shader, and a tuple of (TessCoord.x, TessCoord.y, TessCoord.z, ~) when inside a Tess Eval
         // shader.
         TessCoordInstanceIDVertexID = 47,
+        TexCoord_0 = 48,
+        TexCoord_7 = 55,
         // This attribute contains a tuple of (Unk, Unk, Unk, gl_FrontFacing) when inside a fragment
         // shader. It is unknown what the other values contain.
         FrontFacing = 63,
@@ -911,14 +917,9 @@ union Instruction {
     } fadd32i;
 
     union {
-        BitField<20, 8, u64> shift_position;
-        BitField<28, 8, u64> shift_length;
-        BitField<48, 1, u64> negate_b;
-        BitField<49, 1, u64> negate_a;
-
-        u64 GetLeftShiftValue() const {
-            return 32 - (shift_position + shift_length);
-        }
+        BitField<40, 1, u64> brev;
+        BitField<47, 1, u64> rd_cc;
+        BitField<48, 1, u64> is_signed;
     } bfe;
 
     union {
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 7d7137109..e8f763ce9 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -140,71 +140,6 @@ void GPU::FlushCommands() {
     renderer.Rasterizer().FlushCommands();
 }
 
-u32 RenderTargetBytesPerPixel(RenderTargetFormat format) {
-    ASSERT(format != RenderTargetFormat::NONE);
-
-    switch (format) {
-    case RenderTargetFormat::RGBA32_FLOAT:
-    case RenderTargetFormat::RGBA32_UINT:
-        return 16;
-    case RenderTargetFormat::RGBA16_UINT:
-    case RenderTargetFormat::RGBA16_UNORM:
-    case RenderTargetFormat::RGBA16_FLOAT:
-    case RenderTargetFormat::RGBX16_FLOAT:
-    case RenderTargetFormat::RG32_FLOAT:
-    case RenderTargetFormat::RG32_UINT:
-        return 8;
-    case RenderTargetFormat::RGBA8_UNORM:
-    case RenderTargetFormat::RGBA8_SNORM:
-    case RenderTargetFormat::RGBA8_SRGB:
-    case RenderTargetFormat::RGBA8_UINT:
-    case RenderTargetFormat::RGB10_A2_UNORM:
-    case RenderTargetFormat::BGRA8_UNORM:
-    case RenderTargetFormat::BGRA8_SRGB:
-    case RenderTargetFormat::RG16_UNORM:
-    case RenderTargetFormat::RG16_SNORM:
-    case RenderTargetFormat::RG16_UINT:
-    case RenderTargetFormat::RG16_SINT:
-    case RenderTargetFormat::RG16_FLOAT:
-    case RenderTargetFormat::R32_FLOAT:
-    case RenderTargetFormat::R11G11B10_FLOAT:
-    case RenderTargetFormat::R32_UINT:
-        return 4;
-    case RenderTargetFormat::R16_UNORM:
-    case RenderTargetFormat::R16_SNORM:
-    case RenderTargetFormat::R16_UINT:
-    case RenderTargetFormat::R16_SINT:
-    case RenderTargetFormat::R16_FLOAT:
-    case RenderTargetFormat::RG8_UNORM:
-    case RenderTargetFormat::RG8_SNORM:
-        return 2;
-    case RenderTargetFormat::R8_UNORM:
-    case RenderTargetFormat::R8_UINT:
-        return 1;
-    default:
-        UNIMPLEMENTED_MSG("Unimplemented render target format {}", static_cast<u32>(format));
-        return 1;
-    }
-}
-
-u32 DepthFormatBytesPerPixel(DepthFormat format) {
-    switch (format) {
-    case DepthFormat::Z32_S8_X24_FLOAT:
-        return 8;
-    case DepthFormat::Z32_FLOAT:
-    case DepthFormat::S8_Z24_UNORM:
-    case DepthFormat::Z24_X8_UNORM:
-    case DepthFormat::Z24_S8_UNORM:
-    case DepthFormat::Z24_C8_UNORM:
-        return 4;
-    case DepthFormat::Z16_UNORM:
-        return 2;
-    default:
-        UNIMPLEMENTED_MSG("Unimplemented Depth format {}", static_cast<u32>(format));
-        return 1;
-    }
-}
-
 // Note that, traditionally, methods are treated as 4-byte addressable locations, and hence
 // their numbers are written down multiplied by 4 in Docs. Here we are not multiply by 4.
 // So the values you see in docs might be multiplied by 4.
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 07727210c..64acb17df 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -39,6 +39,7 @@ enum class RenderTargetFormat : u32 {
     RGBA32_FLOAT = 0xC0,
     RGBA32_UINT = 0xC2,
     RGBA16_UNORM = 0xC6,
+    RGBA16_SNORM = 0xC7,
     RGBA16_UINT = 0xC9,
     RGBA16_FLOAT = 0xCA,
     RG32_FLOAT = 0xCB,
@@ -57,6 +58,7 @@ enum class RenderTargetFormat : u32 {
     RG16_UINT = 0xDD,
     RG16_FLOAT = 0xDE,
     R11G11B10_FLOAT = 0xE0,
+    R32_SINT = 0xE3,
     R32_UINT = 0xE4,
     R32_FLOAT = 0xE5,
     B5G6R5_UNORM = 0xE8,
@@ -82,12 +84,6 @@ enum class DepthFormat : u32 {
     Z32_S8_X24_FLOAT = 0x19,
 };
 
-/// Returns the number of bytes per pixel of each rendertarget format.
-u32 RenderTargetBytesPerPixel(RenderTargetFormat format);
-
-/// Returns the number of bytes per pixel of each depth format.
-u32 DepthFormatBytesPerPixel(DepthFormat format);
-
 struct CommandListHeader;
 class DebugContext;
 
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index 2cdf1aa7f..b1088af3d 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -5,7 +5,7 @@
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "core/core.h"
-#include "core/frontend/scope_acquire_window_context.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/gpu.h"
 #include "video_core/gpu_thread.h"
@@ -27,7 +27,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
         return;
     }
 
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{renderer.GetRenderWindow()};
+    Core::Frontend::ScopeAcquireContext acquire_context{renderer.GetRenderWindow()};
 
     CommandDataContainer next;
     while (state.is_running) {
diff --git a/src/video_core/guest_driver.cpp b/src/video_core/guest_driver.cpp
index 6adef459e..f058f2744 100644
--- a/src/video_core/guest_driver.cpp
+++ b/src/video_core/guest_driver.cpp
@@ -4,13 +4,15 @@
 
 #include <algorithm>
 #include <limits>
+#include <vector>
 
+#include "common/common_types.h"
 #include "video_core/guest_driver.h"
 
 namespace VideoCore {
 
-void GuestDriverProfile::DeduceTextureHandlerSize(std::vector<u32>&& bound_offsets) {
-    if (texture_handler_size_deduced) {
+void GuestDriverProfile::DeduceTextureHandlerSize(std::vector<u32> bound_offsets) {
+    if (texture_handler_size) {
         return;
     }
     const std::size_t size = bound_offsets.size();
@@ -29,7 +31,6 @@ void GuestDriverProfile::DeduceTextureHandlerSize(std::vector<u32>&& bound_offse
     if (min_val > 2) {
         return;
     }
-    texture_handler_size_deduced = true;
     texture_handler_size = min_texture_handler_size * min_val;
 }
 
diff --git a/src/video_core/guest_driver.h b/src/video_core/guest_driver.h
index fc1917347..99450777e 100644
--- a/src/video_core/guest_driver.h
+++ b/src/video_core/guest_driver.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <optional>
 #include <vector>
 
 #include "common/common_types.h"
@@ -17,25 +18,29 @@ namespace VideoCore {
  */
 class GuestDriverProfile {
 public:
-    void DeduceTextureHandlerSize(std::vector<u32>&& bound_offsets);
+    explicit GuestDriverProfile() = default;
+    explicit GuestDriverProfile(std::optional<u32> texture_handler_size)
+        : texture_handler_size{texture_handler_size} {}
+
+    void DeduceTextureHandlerSize(std::vector<u32> bound_offsets);
 
     u32 GetTextureHandlerSize() const {
-        return texture_handler_size;
+        return texture_handler_size.value_or(default_texture_handler_size);
     }
 
-    bool TextureHandlerSizeKnown() const {
-        return texture_handler_size_deduced;
+    bool IsTextureHandlerSizeKnown() const {
+        return texture_handler_size.has_value();
     }
 
 private:
     // Minimum size of texture handler any driver can use.
     static constexpr u32 min_texture_handler_size = 4;
-    // This goes with Vulkan and OpenGL standards but Nvidia GPUs can easily
-    // use 4 bytes instead. Thus, certain drivers may squish the size.
+
+    // This goes with Vulkan and OpenGL standards but Nvidia GPUs can easily use 4 bytes instead.
+    // Thus, certain drivers may squish the size.
     static constexpr u32 default_texture_handler_size = 8;
 
-    u32 texture_handler_size = default_texture_handler_size;
-    bool texture_handler_size_deduced = false;
+    std::optional<u32> texture_handler_size = default_texture_handler_size;
 };
 
 } // namespace VideoCore
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index aea010087..073bdb491 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -174,7 +174,7 @@ private:
     /// End of address space, based on address space in bits.
     static constexpr GPUVAddr address_space_end{1ULL << address_space_width};
 
-    Common::PageTable page_table{page_bits};
+    Common::BackingPageTable page_table{page_bits};
     VMAMap vma_map;
     VideoCore::RasterizerInterface& rasterizer;
 
diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp
index 2f2fe6859..6d522c318 100644
--- a/src/video_core/morton.cpp
+++ b/src/video_core/morton.cpp
@@ -51,6 +51,7 @@ static constexpr ConversionArray morton_to_linear_fns = {
     MortonCopy<true, PixelFormat::R8UI>,
     MortonCopy<true, PixelFormat::RGBA16F>,
     MortonCopy<true, PixelFormat::RGBA16U>,
+    MortonCopy<true, PixelFormat::RGBA16S>,
     MortonCopy<true, PixelFormat::RGBA16UI>,
     MortonCopy<true, PixelFormat::R11FG11FB10F>,
     MortonCopy<true, PixelFormat::RGBA32UI>,
@@ -85,6 +86,7 @@ static constexpr ConversionArray morton_to_linear_fns = {
     MortonCopy<true, PixelFormat::RG32UI>,
     MortonCopy<true, PixelFormat::RGBX16F>,
     MortonCopy<true, PixelFormat::R32UI>,
+    MortonCopy<true, PixelFormat::R32I>,
     MortonCopy<true, PixelFormat::ASTC_2D_8X8>,
     MortonCopy<true, PixelFormat::ASTC_2D_8X5>,
     MortonCopy<true, PixelFormat::ASTC_2D_5X4>,
@@ -130,6 +132,7 @@ static constexpr ConversionArray linear_to_morton_fns = {
     MortonCopy<false, PixelFormat::R8U>,
     MortonCopy<false, PixelFormat::R8UI>,
     MortonCopy<false, PixelFormat::RGBA16F>,
+    MortonCopy<false, PixelFormat::RGBA16S>,
     MortonCopy<false, PixelFormat::RGBA16U>,
     MortonCopy<false, PixelFormat::RGBA16UI>,
     MortonCopy<false, PixelFormat::R11FG11FB10F>,
@@ -166,6 +169,7 @@ static constexpr ConversionArray linear_to_morton_fns = {
     MortonCopy<false, PixelFormat::RG32UI>,
     MortonCopy<false, PixelFormat::RGBX16F>,
     MortonCopy<false, PixelFormat::R32UI>,
+    MortonCopy<false, PixelFormat::R32I>,
     nullptr,
     nullptr,
     nullptr,
diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h
index f18eaf4bc..1a68e3caa 100644
--- a/src/video_core/rasterizer_interface.h
+++ b/src/video_core/rasterizer_interface.h
@@ -25,7 +25,6 @@ constexpr std::size_t NumQueryTypes = 1;
 
 enum class LoadCallbackStage {
     Prepare,
-    Decompile,
     Build,
     Complete,
 };
@@ -89,6 +88,9 @@ public:
     virtual void LoadDiskResources(const std::atomic_bool& stop_loading = false,
                                    const DiskResourceLoadCallback& callback = {}) {}
 
+    /// Initializes renderer dirty flags
+    virtual void SetupDirtyFlags() {}
+
     /// Grant access to the Guest Driver Profile for recording/obtaining info on the guest driver.
     GuestDriverProfile& AccessGuestDriverProfile() {
         return guest_driver_profile;
diff --git a/src/video_core/renderer_base.h b/src/video_core/renderer_base.h
index af1bebc4f..5ec99a126 100644
--- a/src/video_core/renderer_base.h
+++ b/src/video_core/renderer_base.h
@@ -35,15 +35,19 @@ public:
     explicit RendererBase(Core::Frontend::EmuWindow& window);
     virtual ~RendererBase();
 
-    /// Swap buffers (render frame)
-    virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
-
     /// Initialize the renderer
     virtual bool Init() = 0;
 
     /// Shutdown the renderer
     virtual void ShutDown() = 0;
 
+    /// Finalize rendering the guest frame and draw into the presentation texture
+    virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
+
+    /// Draws the latest frame to the window waiting timeout_ms for a frame to arrive (Renderer
+    /// specific implementation)
+    virtual void TryPresent(int timeout_ms) = 0;
+
     // Getter/setter functions:
     // ------------------------
 
diff --git a/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp b/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp
index 874ed3c6e..b8a512cb6 100644
--- a/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_framebuffer_cache.cpp
@@ -11,7 +11,6 @@
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_framebuffer_cache.h"
-#include "video_core/renderer_opengl/gl_state.h"
 
 namespace OpenGL {
 
@@ -36,8 +35,7 @@ OGLFramebuffer FramebufferCacheOpenGL::CreateFramebuffer(const FramebufferCacheK
     framebuffer.Create();
 
     // TODO(Rodrigo): Use DSA here after Nvidia fixes their framebuffer DSA bugs.
-    local_state.draw.draw_framebuffer = framebuffer.handle;
-    local_state.ApplyFramebufferState();
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer.handle);
 
     if (key.zeta) {
         const bool stencil = key.zeta->GetSurfaceParams().type == SurfaceType::DepthStencil;
diff --git a/src/video_core/renderer_opengl/gl_framebuffer_cache.h b/src/video_core/renderer_opengl/gl_framebuffer_cache.h
index 02ec80ae9..8f698fee0 100644
--- a/src/video_core/renderer_opengl/gl_framebuffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_framebuffer_cache.h
@@ -13,7 +13,6 @@
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 
 namespace OpenGL {
@@ -63,7 +62,6 @@ public:
 private:
     OGLFramebuffer CreateFramebuffer(const FramebufferCacheKey& key);
 
-    OpenGLState local_state;
     std::unordered_map<FramebufferCacheKey, OGLFramebuffer> cache;
 };
 
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index e1965fb21..826eee7df 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -28,7 +28,6 @@
 #include "video_core/renderer_opengl/gl_query_cache.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"
 
@@ -36,6 +35,7 @@ namespace OpenGL {
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
+using Tegra::Engines::ShaderType;
 using VideoCore::Surface::PixelFormat;
 using VideoCore::Surface::SurfaceTarget;
 using VideoCore::Surface::SurfaceType;
@@ -54,10 +54,11 @@ MICROPROFILE_DEFINE(OpenGL_PrimitiveAssembly, "OpenGL", "Prim Asmbl", MP_RGB(255
 
 namespace {
 
+constexpr std::size_t NumSupportedVertexAttributes = 16;
+
 template <typename Engine, typename Entry>
 Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
-                                               Tegra::Engines::ShaderType shader_type,
-                                               std::size_t index = 0) {
+                                               ShaderType shader_type, std::size_t index = 0) {
     if (entry.IsBindless()) {
         const Tegra::Texture::TextureHandle tex_handle =
             engine.AccessConstBuffer32(shader_type, entry.GetBuffer(), entry.GetOffset());
@@ -74,7 +75,7 @@ Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry
 }
 
 std::size_t GetConstBufferSize(const Tegra::Engines::ConstBufferInfo& buffer,
-                               const GLShader::ConstBufferEntry& entry) {
+                               const ConstBufferEntry& entry) {
     if (!entry.IsIndirect()) {
         return entry.GetSize();
     }
@@ -88,18 +89,19 @@ std::size_t GetConstBufferSize(const Tegra::Engines::ConstBufferInfo& buffer,
     return buffer.size;
 }
 
+void oglEnable(GLenum cap, bool state) {
+    (state ? glEnable : glDisable)(cap);
+}
+
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
-                                   ScreenInfo& info)
-    : RasterizerAccelerated{system.Memory()}, texture_cache{system, *this, device},
+                                   ScreenInfo& info, GLShader::ProgramManager& program_manager,
+                                   StateTracker& state_tracker)
+    : RasterizerAccelerated{system.Memory()}, texture_cache{system, *this, device, state_tracker},
       shader_cache{*this, system, emu_window, device}, query_cache{system, *this}, system{system},
-      screen_info{info}, buffer_cache{*this, system, device, STREAM_BUFFER_SIZE} {
-    shader_program_manager = std::make_unique<GLShader::ProgramManager>();
-    state.draw.shader_program = 0;
-    state.Apply();
-
-    LOG_DEBUG(Render_OpenGL, "Sync fixed function OpenGL state here");
+      screen_info{info}, program_manager{program_manager}, state_tracker{state_tracker},
+      buffer_cache{*this, system, device, STREAM_BUFFER_SIZE} {
     CheckExtensions();
 }
 
@@ -113,93 +115,72 @@ void RasterizerOpenGL::CheckExtensions() {
     }
 }
 
-GLuint RasterizerOpenGL::SetupVertexFormat() {
+void RasterizerOpenGL::SetupVertexFormat() {
     auto& gpu = system.GPU().Maxwell3D();
-    const auto& regs = gpu.regs;
-
-    if (!gpu.dirty.vertex_attrib_format) {
-        return state.draw.vertex_array;
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::VertexFormats]) {
+        return;
     }
-    gpu.dirty.vertex_attrib_format = false;
+    flags[Dirty::VertexFormats] = false;
 
     MICROPROFILE_SCOPE(OpenGL_VAO);
 
-    auto [iter, is_cache_miss] = vertex_array_cache.try_emplace(regs.vertex_attrib_format);
-    auto& vao_entry = iter->second;
-
-    if (is_cache_miss) {
-        vao_entry.Create();
-        const GLuint vao = vao_entry.handle;
-
-        // Eventhough we are using DSA to create this vertex array, there is a bug on Intel's blob
-        // that fails to properly create the vertex array if it's not bound even after creating it
-        // with glCreateVertexArrays
-        state.draw.vertex_array = vao;
-        state.ApplyVertexArrayState();
-
-        // Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL.
-        // Enables the first 16 vertex attributes always, as we don't know which ones are actually
-        // used until shader time. Note, Tegra technically supports 32, but we're capping this to 16
-        // for now to avoid OpenGL errors.
-        // TODO(Subv): Analyze the shader to identify which attributes are actually used and don't
-        // assume every shader uses them all.
-        for (u32 index = 0; index < 16; ++index) {
-            const auto& attrib = regs.vertex_attrib_format[index];
-
-            // Ignore invalid attributes.
-            if (!attrib.IsValid())
-                continue;
-
-            const auto& buffer = regs.vertex_array[attrib.buffer];
-            LOG_TRACE(Render_OpenGL,
-                      "vertex attrib {}, count={}, size={}, type={}, offset={}, normalize={}",
-                      index, attrib.ComponentCount(), attrib.SizeString(), attrib.TypeString(),
-                      attrib.offset.Value(), attrib.IsNormalized());
-
-            ASSERT(buffer.IsEnabled());
-
-            glEnableVertexArrayAttrib(vao, index);
-            if (attrib.type == Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::SignedInt ||
-                attrib.type ==
-                    Tegra::Engines::Maxwell3D::Regs::VertexAttribute::Type::UnsignedInt) {
-                glVertexArrayAttribIFormat(vao, index, attrib.ComponentCount(),
-                                           MaxwellToGL::VertexType(attrib), attrib.offset);
-            } else {
-                glVertexArrayAttribFormat(
-                    vao, index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
-                    attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
-            }
-            glVertexArrayAttribBinding(vao, index, attrib.buffer);
+    // Use the vertex array as-is, assumes that the data is formatted correctly for OpenGL. Enables
+    // the first 16 vertex attributes always, as we don't know which ones are actually used until
+    // shader time. Note, Tegra technically supports 32, but we're capping this to 16 for now to
+    // avoid OpenGL errors.
+    // TODO(Subv): Analyze the shader to identify which attributes are actually used and don't
+    // assume every shader uses them all.
+    for (std::size_t index = 0; index < NumSupportedVertexAttributes; ++index) {
+        if (!flags[Dirty::VertexFormat0 + index]) {
+            continue;
         }
-    }
+        flags[Dirty::VertexFormat0 + index] = false;
+
+        const auto attrib = gpu.regs.vertex_attrib_format[index];
+        const auto gl_index = static_cast<GLuint>(index);
 
-    // Rebinding the VAO invalidates the vertex buffer bindings.
-    gpu.dirty.ResetVertexArrays();
+        // Ignore invalid attributes.
+        if (!attrib.IsValid()) {
+            glDisableVertexAttribArray(gl_index);
+            continue;
+        }
+        glEnableVertexAttribArray(gl_index);
 
-    state.draw.vertex_array = vao_entry.handle;
-    return vao_entry.handle;
+        if (attrib.type == Maxwell::VertexAttribute::Type::SignedInt ||
+            attrib.type == Maxwell::VertexAttribute::Type::UnsignedInt) {
+            glVertexAttribIFormat(gl_index, attrib.ComponentCount(),
+                                  MaxwellToGL::VertexType(attrib), attrib.offset);
+        } else {
+            glVertexAttribFormat(gl_index, attrib.ComponentCount(), MaxwellToGL::VertexType(attrib),
+                                 attrib.IsNormalized() ? GL_TRUE : GL_FALSE, attrib.offset);
+        }
+        glVertexAttribBinding(gl_index, attrib.buffer);
+    }
 }
 
-void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
+void RasterizerOpenGL::SetupVertexBuffer() {
     auto& gpu = system.GPU().Maxwell3D();
-    if (!gpu.dirty.vertex_array_buffers)
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::VertexBuffers]) {
         return;
-    gpu.dirty.vertex_array_buffers = false;
-
-    const auto& regs = gpu.regs;
+    }
+    flags[Dirty::VertexBuffers] = false;
 
     MICROPROFILE_SCOPE(OpenGL_VB);
 
     // Upload all guest vertex arrays sequentially to our buffer
-    for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
-        if (!gpu.dirty.vertex_array[index])
+    const auto& regs = gpu.regs;
+    for (std::size_t index = 0; index < Maxwell::NumVertexArrays; ++index) {
+        if (!flags[Dirty::VertexBuffer0 + index]) {
             continue;
-        gpu.dirty.vertex_array[index] = false;
-        gpu.dirty.vertex_instance[index] = false;
+        }
+        flags[Dirty::VertexBuffer0 + index] = false;
 
         const auto& vertex_array = regs.vertex_array[index];
-        if (!vertex_array.IsEnabled())
+        if (!vertex_array.IsEnabled()) {
             continue;
+        }
 
         const GPUVAddr start = vertex_array.StartAddress();
         const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
@@ -209,42 +190,30 @@ void RasterizerOpenGL::SetupVertexBuffer(GLuint vao) {
         const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
 
         // Bind the vertex array to the buffer at the current offset.
-        vertex_array_pushbuffer.SetVertexBuffer(index, vertex_buffer, vertex_buffer_offset,
-                                                vertex_array.stride);
-
-        if (regs.instanced_arrays.IsInstancingEnabled(index) && vertex_array.divisor != 0) {
-            // Enable vertex buffer instancing with the specified divisor.
-            glVertexArrayBindingDivisor(vao, index, vertex_array.divisor);
-        } else {
-            // Disable the vertex buffer instancing.
-            glVertexArrayBindingDivisor(vao, index, 0);
-        }
+        vertex_array_pushbuffer.SetVertexBuffer(static_cast<GLuint>(index), vertex_buffer,
+                                                vertex_buffer_offset, vertex_array.stride);
     }
 }
 
-void RasterizerOpenGL::SetupVertexInstances(GLuint vao) {
+void RasterizerOpenGL::SetupVertexInstances() {
     auto& gpu = system.GPU().Maxwell3D();
-
-    if (!gpu.dirty.vertex_instances)
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::VertexInstances]) {
         return;
-    gpu.dirty.vertex_instances = false;
+    }
+    flags[Dirty::VertexInstances] = false;
 
     const auto& regs = gpu.regs;
-    // Upload all guest vertex arrays sequentially to our buffer
-    for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
-        if (!gpu.dirty.vertex_instance[index])
+    for (std::size_t index = 0; index < NumSupportedVertexAttributes; ++index) {
+        if (!flags[Dirty::VertexInstance0 + index]) {
             continue;
-
-        gpu.dirty.vertex_instance[index] = false;
-
-        if (regs.instanced_arrays.IsInstancingEnabled(index) &&
-            regs.vertex_array[index].divisor != 0) {
-            // Enable vertex buffer instancing with the specified divisor.
-            glVertexArrayBindingDivisor(vao, index, regs.vertex_array[index].divisor);
-        } else {
-            // Disable the vertex buffer instancing.
-            glVertexArrayBindingDivisor(vao, index, 0);
         }
+        flags[Dirty::VertexInstance0 + index] = false;
+
+        const auto gl_index = static_cast<GLuint>(index);
+        const bool instancing_enabled = regs.instanced_arrays.IsInstancingEnabled(gl_index);
+        const GLuint divisor = instancing_enabled ? regs.vertex_array[index].divisor : 0;
+        glVertexBindingDivisor(gl_index, divisor);
     }
 }
 
@@ -260,8 +229,7 @@ GLintptr RasterizerOpenGL::SetupIndexBuffer() {
 void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     MICROPROFILE_SCOPE(OpenGL_Shader);
     auto& gpu = system.GPU().Maxwell3D();
-
-    std::array<bool, Maxwell::NumClipDistances> clip_distances{};
+    u32 clip_distances = 0;
 
     for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
         const auto& shader_config = gpu.regs.shader_config[index];
@@ -271,10 +239,10 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
         if (!gpu.regs.IsShaderConfigEnabled(index)) {
             switch (program) {
             case Maxwell::ShaderProgram::Geometry:
-                shader_program_manager->UseTrivialGeometryShader();
+                program_manager.UseGeometryShader(0);
                 break;
             case Maxwell::ShaderProgram::Fragment:
-                shader_program_manager->UseTrivialFragmentShader();
+                program_manager.UseFragmentShader(0);
                 break;
             default:
                 break;
@@ -299,19 +267,17 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
         SetupDrawTextures(stage, shader);
         SetupDrawImages(stage, shader);
 
-        const ProgramVariant variant(primitive_mode);
-        const auto program_handle = shader->GetHandle(variant);
-
+        const GLuint program_handle = shader->GetHandle();
         switch (program) {
         case Maxwell::ShaderProgram::VertexA:
         case Maxwell::ShaderProgram::VertexB:
-            shader_program_manager->UseProgrammableVertexShader(program_handle);
+            program_manager.UseVertexShader(program_handle);
             break;
         case Maxwell::ShaderProgram::Geometry:
-            shader_program_manager->UseProgrammableGeometryShader(program_handle);
+            program_manager.UseGeometryShader(program_handle);
             break;
         case Maxwell::ShaderProgram::Fragment:
-            shader_program_manager->UseProgrammableFragmentShader(program_handle);
+            program_manager.UseFragmentShader(program_handle);
             break;
         default:
             UNIMPLEMENTED_MSG("Unimplemented shader index={}, enable={}, offset=0x{:08X}", index,
@@ -322,9 +288,7 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
         // When a clip distance is enabled but not set in the shader it crops parts of the screen
         // (sometimes it's half the screen, sometimes three quarters). To avoid this, enable the
         // clip distances only when it's written by a shader stage.
-        for (std::size_t i = 0; i < Maxwell::NumClipDistances; ++i) {
-            clip_distances[i] = clip_distances[i] || shader->GetShaderEntries().clip_distances[i];
-        }
+        clip_distances |= shader->GetEntries().clip_distances;
 
         // When VertexA is enabled, we have dual vertex shaders
         if (program == Maxwell::ShaderProgram::VertexA) {
@@ -334,8 +298,7 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     }
 
     SyncClipEnabled(clip_distances);
-
-    gpu.dirty.shaders = false;
+    gpu.dirty.flags[Dirty::Shaders] = false;
 }
 
 std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
@@ -368,20 +331,23 @@ void RasterizerOpenGL::LoadDiskResources(const std::atomic_bool& stop_loading,
     shader_cache.LoadDiskCache(stop_loading, callback);
 }
 
+void RasterizerOpenGL::SetupDirtyFlags() {
+    state_tracker.Initialize();
+}
+
 void RasterizerOpenGL::ConfigureFramebuffers() {
     MICROPROFILE_SCOPE(OpenGL_Framebuffer);
     auto& gpu = system.GPU().Maxwell3D();
-    if (!gpu.dirty.render_settings) {
+    if (!gpu.dirty.flags[VideoCommon::Dirty::RenderTargets]) {
         return;
     }
-    gpu.dirty.render_settings = false;
+    gpu.dirty.flags[VideoCommon::Dirty::RenderTargets] = false;
 
     texture_cache.GuardRenderTargets(true);
 
     View depth_surface = texture_cache.GetDepthBufferSurface(true);
 
     const auto& regs = gpu.regs;
-    state.framebuffer_srgb.enabled = regs.framebuffer_srgb != 0;
     UNIMPLEMENTED_IF(regs.rt_separate_frag_data == 0);
 
     // Bind the framebuffer surfaces
@@ -409,14 +375,11 @@ void RasterizerOpenGL::ConfigureFramebuffers() {
 
     texture_cache.GuardRenderTargets(false);
 
-    state.draw.draw_framebuffer = framebuffer_cache.GetFramebuffer(key);
-    SyncViewport(state);
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer_cache.GetFramebuffer(key));
 }
 
-void RasterizerOpenGL::ConfigureClearFramebuffer(OpenGLState& current_state, bool using_color_fb,
-                                                 bool using_depth_fb, bool using_stencil_fb) {
-    using VideoCore::Surface::SurfaceType;
-
+void RasterizerOpenGL::ConfigureClearFramebuffer(bool using_color_fb, bool using_depth_fb,
+                                                 bool using_stencil_fb) {
     auto& gpu = system.GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
@@ -435,80 +398,44 @@ void RasterizerOpenGL::ConfigureClearFramebuffer(OpenGLState& current_state, boo
     key.colors[0] = color_surface;
     key.zeta = depth_surface;
 
-    current_state.draw.draw_framebuffer = framebuffer_cache.GetFramebuffer(key);
-    current_state.ApplyFramebufferState();
+    state_tracker.NotifyFramebuffer();
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, framebuffer_cache.GetFramebuffer(key));
 }
 
 void RasterizerOpenGL::Clear() {
-    const auto& maxwell3d = system.GPU().Maxwell3D();
-
-    if (!maxwell3d.ShouldExecute()) {
+    const auto& gpu = system.GPU().Maxwell3D();
+    if (!gpu.ShouldExecute()) {
         return;
     }
 
-    const auto& regs = maxwell3d.regs;
+    const auto& regs = gpu.regs;
     bool use_color{};
     bool use_depth{};
     bool use_stencil{};
 
-    OpenGLState prev_state{OpenGLState::GetCurState()};
-    SCOPE_EXIT({
-        prev_state.AllDirty();
-        prev_state.Apply();
-    });
-
-    OpenGLState clear_state{OpenGLState::GetCurState()};
-    clear_state.SetDefaultViewports();
     if (regs.clear_buffers.R || regs.clear_buffers.G || regs.clear_buffers.B ||
         regs.clear_buffers.A) {
         use_color = true;
     }
     if (use_color) {
-        clear_state.color_mask[0].red_enabled = regs.clear_buffers.R ? GL_TRUE : GL_FALSE;
-        clear_state.color_mask[0].green_enabled = regs.clear_buffers.G ? GL_TRUE : GL_FALSE;
-        clear_state.color_mask[0].blue_enabled = regs.clear_buffers.B ? GL_TRUE : GL_FALSE;
-        clear_state.color_mask[0].alpha_enabled = regs.clear_buffers.A ? GL_TRUE : GL_FALSE;
+        state_tracker.NotifyColorMask0();
+        glColorMaski(0, regs.clear_buffers.R != 0, regs.clear_buffers.G != 0,
+                     regs.clear_buffers.B != 0, regs.clear_buffers.A != 0);
+
+        // TODO(Rodrigo): Determine if clamping is used on clears
+        SyncFragmentColorClampState();
+        SyncFramebufferSRGB();
     }
     if (regs.clear_buffers.Z) {
         ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear Z but buffer is not enabled!");
         use_depth = true;
 
-        // Always enable the depth write when clearing the depth buffer. The depth write mask is
-        // ignored when clearing the buffer in the Switch, but OpenGL obeys it so we set it to
-        // true.
-        clear_state.depth.test_enabled = true;
-        clear_state.depth.test_func = GL_ALWAYS;
-        clear_state.depth.write_mask = GL_TRUE;
+        state_tracker.NotifyDepthMask();
+        glDepthMask(GL_TRUE);
     }
     if (regs.clear_buffers.S) {
-        ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear stencil but buffer is not enabled!");
+        ASSERT_MSG(regs.zeta_enable, "Tried to clear stencil but buffer is not enabled!");
         use_stencil = true;
-        clear_state.stencil.test_enabled = true;
-
-        if (regs.clear_flags.stencil) {
-            // Stencil affects the clear so fill it with the used masks
-            clear_state.stencil.front.test_func = GL_ALWAYS;
-            clear_state.stencil.front.test_mask = regs.stencil_front_func_mask;
-            clear_state.stencil.front.action_stencil_fail = GL_KEEP;
-            clear_state.stencil.front.action_depth_fail = GL_KEEP;
-            clear_state.stencil.front.action_depth_pass = GL_KEEP;
-            clear_state.stencil.front.write_mask = regs.stencil_front_mask;
-            if (regs.stencil_two_side_enable) {
-                clear_state.stencil.back.test_func = GL_ALWAYS;
-                clear_state.stencil.back.test_mask = regs.stencil_back_func_mask;
-                clear_state.stencil.back.action_stencil_fail = GL_KEEP;
-                clear_state.stencil.back.action_depth_fail = GL_KEEP;
-                clear_state.stencil.back.action_depth_pass = GL_KEEP;
-                clear_state.stencil.back.write_mask = regs.stencil_back_mask;
-            } else {
-                clear_state.stencil.back.test_func = GL_ALWAYS;
-                clear_state.stencil.back.test_mask = 0xFFFFFFFF;
-                clear_state.stencil.back.write_mask = 0xFFFFFFFF;
-                clear_state.stencil.back.action_stencil_fail = GL_KEEP;
-                clear_state.stencil.back.action_depth_fail = GL_KEEP;
-                clear_state.stencil.back.action_depth_pass = GL_KEEP;
-            }
-        }
     }
 
     if (!use_color && !use_depth && !use_stencil) {
@@ -516,20 +443,18 @@ void RasterizerOpenGL::Clear() {
         return;
     }
 
-    ConfigureClearFramebuffer(clear_state, use_color, use_depth, use_stencil);
+    SyncRasterizeEnable();
 
-    SyncViewport(clear_state);
-    SyncRasterizeEnable(clear_state);
     if (regs.clear_flags.scissor) {
-        SyncScissorTest(clear_state);
+        SyncScissorTest();
+    } else {
+        state_tracker.NotifyScissor0();
+        glDisablei(GL_SCISSOR_TEST, 0);
     }
 
-    if (regs.clear_flags.viewport) {
-        clear_state.EmulateViewportWithScissor();
-    }
+    UNIMPLEMENTED_IF(regs.clear_flags.viewport);
 
-    clear_state.AllDirty();
-    clear_state.Apply();
+    ConfigureClearFramebuffer(use_color, use_depth, use_stencil);
 
     if (use_color) {
         glClearBufferfv(GL_COLOR, 0, regs.clear_color);
@@ -549,25 +474,27 @@ void RasterizerOpenGL::Clear() {
 void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
     auto& gpu = system.GPU().Maxwell3D();
-    const auto& regs = gpu.regs;
 
     query_cache.UpdateCounters();
 
-    SyncRasterizeEnable(state);
+    SyncViewport();
+    SyncRasterizeEnable();
+    SyncPolygonModes();
     SyncColorMask();
     SyncFragmentColorClampState();
     SyncMultiSampleState();
     SyncDepthTestState();
+    SyncDepthClamp();
     SyncStencilTestState();
     SyncBlendState();
     SyncLogicOpState();
     SyncCullMode();
     SyncPrimitiveRestart();
-    SyncScissorTest(state);
-    SyncTransformFeedback();
+    SyncScissorTest();
     SyncPointState();
     SyncPolygonOffset();
     SyncAlphaTest();
+    SyncFramebufferSRGB();
 
     buffer_cache.Acquire();
 
@@ -591,14 +518,13 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     buffer_cache.Map(buffer_size);
 
     // Prepare vertex array format.
-    const GLuint vao = SetupVertexFormat();
-    vertex_array_pushbuffer.Setup(vao);
+    SetupVertexFormat();
+    vertex_array_pushbuffer.Setup();
 
     // Upload vertex and index data.
-    SetupVertexBuffer(vao);
-    SetupVertexInstances(vao);
-
-    GLintptr index_buffer_offset;
+    SetupVertexBuffer();
+    SetupVertexInstances();
+    GLintptr index_buffer_offset = 0;
     if (is_indexed) {
         index_buffer_offset = SetupIndexBuffer();
     }
@@ -624,27 +550,20 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     ConfigureFramebuffers();
 
     // Signal the buffer cache that we are not going to upload more things.
-    const bool invalidate = buffer_cache.Unmap();
+    buffer_cache.Unmap();
 
     // Now that we are no longer uploading data, we can safely bind the buffers to OpenGL.
     vertex_array_pushbuffer.Bind();
     bind_ubo_pushbuffer.Bind();
     bind_ssbo_pushbuffer.Bind();
 
-    if (invalidate) {
-        // As all cached buffers are invalidated, we need to recheck their state.
-        gpu.dirty.ResetVertexArrays();
-    }
-    gpu.dirty.memory_general = false;
-
-    shader_program_manager->ApplyTo(state);
-    state.Apply();
+    program_manager.BindGraphicsPipeline();
 
     if (texture_cache.TextureBarrier()) {
         glTextureBarrier();
     }
 
-    ++num_queued_commands;
+    BeginTransformFeedback(primitive_mode);
 
     const GLuint base_instance = static_cast<GLuint>(gpu.regs.vb_base_instance);
     const GLsizei num_instances =
@@ -683,6 +602,10 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
                                               num_instances, base_instance);
         }
     }
+
+    EndTransformFeedback();
+
+    ++num_queued_commands;
 }
 
 void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
@@ -695,13 +618,7 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     auto kernel = shader_cache.GetComputeKernel(code_addr);
     SetupComputeTextures(kernel);
     SetupComputeImages(kernel);
-
-    const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
-    const ProgramVariant variant(launch_desc.block_dim_x, launch_desc.block_dim_y,
-                                 launch_desc.block_dim_z, launch_desc.shared_alloc,
-                                 launch_desc.local_pos_alloc);
-    state.draw.shader_program = kernel->GetHandle(variant);
-    state.draw.program_pipeline = 0;
+    program_manager.BindComputeShader(kernel->GetHandle());
 
     const std::size_t buffer_size =
         Tegra::Engines::KeplerCompute::NumConstBuffers *
@@ -719,11 +636,7 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     bind_ubo_pushbuffer.Bind();
     bind_ssbo_pushbuffer.Bind();
 
-    state.ApplyTextures();
-    state.ApplyImages();
-    state.ApplyShaderProgram();
-    state.ApplyProgramPipeline();
-
+    const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
     glDispatchCompute(launch_desc.grid_dim_x, launch_desc.grid_dim_y, launch_desc.grid_dim_z);
     ++num_queued_commands;
 }
@@ -828,7 +741,7 @@ void RasterizerOpenGL::SetupDrawConstBuffers(std::size_t stage_index, const Shad
     const auto& shader_stage = stages[stage_index];
 
     u32 binding = device.GetBaseBindings(stage_index).uniform_buffer;
-    for (const auto& entry : shader->GetShaderEntries().const_buffers) {
+    for (const auto& entry : shader->GetEntries().const_buffers) {
         const auto& buffer = shader_stage.const_buffers[entry.GetIndex()];
         SetupConstBuffer(binding++, buffer, entry);
     }
@@ -839,7 +752,7 @@ void RasterizerOpenGL::SetupComputeConstBuffers(const Shader& kernel) {
     const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
 
     u32 binding = 0;
-    for (const auto& entry : kernel->GetShaderEntries().const_buffers) {
+    for (const auto& entry : kernel->GetEntries().const_buffers) {
         const auto& config = launch_desc.const_buffer_config[entry.GetIndex()];
         const std::bitset<8> mask = launch_desc.const_buffer_enable_mask.Value();
         Tegra::Engines::ConstBufferInfo buffer;
@@ -851,7 +764,7 @@ void RasterizerOpenGL::SetupComputeConstBuffers(const Shader& kernel) {
 }
 
 void RasterizerOpenGL::SetupConstBuffer(u32 binding, const Tegra::Engines::ConstBufferInfo& buffer,
-                                        const GLShader::ConstBufferEntry& entry) {
+                                        const ConstBufferEntry& entry) {
     if (!buffer.enabled) {
         // Set values to zero to unbind buffers
         bind_ubo_pushbuffer.Push(binding, buffer_cache.GetEmptyBuffer(sizeof(float)), 0,
@@ -875,7 +788,7 @@ void RasterizerOpenGL::SetupDrawGlobalMemory(std::size_t stage_index, const Shad
     const auto cbufs{gpu.Maxwell3D().state.shader_stages[stage_index]};
 
     u32 binding = device.GetBaseBindings(stage_index).shader_storage_buffer;
-    for (const auto& entry : shader->GetShaderEntries().global_memory_entries) {
+    for (const auto& entry : shader->GetEntries().global_memory_entries) {
         const auto addr{cbufs.const_buffers[entry.GetCbufIndex()].address + entry.GetCbufOffset()};
         const auto gpu_addr{memory_manager.Read<u64>(addr)};
         const auto size{memory_manager.Read<u32>(addr + 8)};
@@ -889,7 +802,7 @@ void RasterizerOpenGL::SetupComputeGlobalMemory(const Shader& kernel) {
     const auto cbufs{gpu.KeplerCompute().launch_description.const_buffer_config};
 
     u32 binding = 0;
-    for (const auto& entry : kernel->GetShaderEntries().global_memory_entries) {
+    for (const auto& entry : kernel->GetEntries().global_memory_entries) {
         const auto addr{cbufs[entry.GetCbufIndex()].Address() + entry.GetCbufOffset()};
         const auto gpu_addr{memory_manager.Read<u64>(addr)};
         const auto size{memory_manager.Read<u32>(addr + 8)};
@@ -897,7 +810,7 @@ void RasterizerOpenGL::SetupComputeGlobalMemory(const Shader& kernel) {
     }
 }
 
-void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GLShader::GlobalMemoryEntry& entry,
+void RasterizerOpenGL::SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry,
                                          GPUVAddr gpu_addr, std::size_t size) {
     const auto alignment{device.GetShaderStorageBufferAlignment()};
     const auto [ssbo, buffer_offset] =
@@ -909,16 +822,11 @@ void RasterizerOpenGL::SetupDrawTextures(std::size_t stage_index, const Shader&
     MICROPROFILE_SCOPE(OpenGL_Texture);
     const auto& maxwell3d = system.GPU().Maxwell3D();
     u32 binding = device.GetBaseBindings(stage_index).sampler;
-    for (const auto& entry : shader->GetShaderEntries().samplers) {
-        const auto shader_type = static_cast<Tegra::Engines::ShaderType>(stage_index);
-        if (!entry.IsIndexed()) {
-            const auto texture = GetTextureInfo(maxwell3d, entry, shader_type);
+    for (const auto& entry : shader->GetEntries().samplers) {
+        const auto shader_type = static_cast<ShaderType>(stage_index);
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
+            const auto texture = GetTextureInfo(maxwell3d, entry, shader_type, i);
             SetupTexture(binding++, texture, entry);
-        } else {
-            for (std::size_t i = 0; i < entry.Size(); ++i) {
-                const auto texture = GetTextureInfo(maxwell3d, entry, shader_type, i);
-                SetupTexture(binding++, texture, entry);
-            }
         }
     }
 }
@@ -927,46 +835,39 @@ void RasterizerOpenGL::SetupComputeTextures(const Shader& kernel) {
     MICROPROFILE_SCOPE(OpenGL_Texture);
     const auto& compute = system.GPU().KeplerCompute();
     u32 binding = 0;
-    for (const auto& entry : kernel->GetShaderEntries().samplers) {
-        if (!entry.IsIndexed()) {
-            const auto texture =
-                GetTextureInfo(compute, entry, Tegra::Engines::ShaderType::Compute);
+    for (const auto& entry : kernel->GetEntries().samplers) {
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
+            const auto texture = GetTextureInfo(compute, entry, ShaderType::Compute, i);
             SetupTexture(binding++, texture, entry);
-        } else {
-            for (std::size_t i = 0; i < entry.Size(); ++i) {
-                const auto texture =
-                    GetTextureInfo(compute, entry, Tegra::Engines::ShaderType::Compute, i);
-                SetupTexture(binding++, texture, entry);
-            }
         }
     }
 }
 
 void RasterizerOpenGL::SetupTexture(u32 binding, const Tegra::Texture::FullTextureInfo& texture,
-                                    const GLShader::SamplerEntry& entry) {
+                                    const SamplerEntry& entry) {
     const auto view = texture_cache.GetTextureSurface(texture.tic, entry);
     if (!view) {
         // Can occur when texture addr is null or its memory is unmapped/invalid
-        state.samplers[binding] = 0;
-        state.textures[binding] = 0;
+        glBindSampler(binding, 0);
+        glBindTextureUnit(binding, 0);
         return;
     }
-    state.textures[binding] = view->GetTexture();
+    glBindTextureUnit(binding, view->GetTexture());
 
     if (view->GetSurfaceParams().IsBuffer()) {
         return;
     }
-    state.samplers[binding] = sampler_cache.GetSampler(texture.tsc);
-
     // Apply swizzle to textures that are not buffers.
     view->ApplySwizzle(texture.tic.x_source, texture.tic.y_source, texture.tic.z_source,
                        texture.tic.w_source);
+
+    glBindSampler(binding, sampler_cache.GetSampler(texture.tsc));
 }
 
 void RasterizerOpenGL::SetupDrawImages(std::size_t stage_index, const Shader& shader) {
     const auto& maxwell3d = system.GPU().Maxwell3D();
     u32 binding = device.GetBaseBindings(stage_index).image;
-    for (const auto& entry : shader->GetShaderEntries().images) {
+    for (const auto& entry : shader->GetEntries().images) {
         const auto shader_type = static_cast<Tegra::Engines::ShaderType>(stage_index);
         const auto tic = GetTextureInfo(maxwell3d, entry, shader_type).tic;
         SetupImage(binding++, tic, entry);
@@ -976,17 +877,17 @@ void RasterizerOpenGL::SetupDrawImages(std::size_t stage_index, const Shader& sh
 void RasterizerOpenGL::SetupComputeImages(const Shader& shader) {
     const auto& compute = system.GPU().KeplerCompute();
     u32 binding = 0;
-    for (const auto& entry : shader->GetShaderEntries().images) {
+    for (const auto& entry : shader->GetEntries().images) {
         const auto tic = GetTextureInfo(compute, entry, Tegra::Engines::ShaderType::Compute).tic;
         SetupImage(binding++, tic, entry);
     }
 }
 
 void RasterizerOpenGL::SetupImage(u32 binding, const Tegra::Texture::TICEntry& tic,
-                                  const GLShader::ImageEntry& entry) {
+                                  const ImageEntry& entry) {
     const auto view = texture_cache.GetImageSurface(tic, entry);
     if (!view) {
-        state.images[binding] = 0;
+        glBindImageTexture(binding, 0, 0, GL_FALSE, 0, GL_READ_ONLY, GL_R8);
         return;
     }
     if (!tic.IsBuffer()) {
@@ -995,55 +896,87 @@ void RasterizerOpenGL::SetupImage(u32 binding, const Tegra::Texture::TICEntry& t
     if (entry.IsWritten()) {
         view->MarkAsModified(texture_cache.Tick());
     }
-    state.images[binding] = view->GetTexture();
+    glBindImageTexture(binding, view->GetTexture(), 0, GL_TRUE, 0, GL_READ_WRITE,
+                       view->GetFormat());
 }
 
-void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
-    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 =
-        geometry_shaders_enabled ? Tegra::Engines::Maxwell3D::Regs::NumViewports : 1;
-    for (std::size_t i = 0; i < viewport_count; i++) {
-        auto& viewport = current_state.viewports[i];
-        const auto& src = regs.viewports[i];
-        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();
-        viewport.height = viewport_rect.GetHeight();
-        viewport.depth_range_far = src.depth_range_far;
-        viewport.depth_range_near = src.depth_range_near;
-    }
-    state.depth_clamp.far_plane = regs.view_volume_clip_control.depth_clamp_far != 0;
-    state.depth_clamp.near_plane = regs.view_volume_clip_control.depth_clamp_near != 0;
-
-    bool flip_y = false;
-    if (regs.viewport_transform[0].scale_y < 0.0) {
-        flip_y = !flip_y;
-    }
-    if (regs.screen_y_control.y_negate != 0) {
-        flip_y = !flip_y;
-    }
-    state.clip_control.origin = flip_y ? GL_UPPER_LEFT : GL_LOWER_LEFT;
-    state.clip_control.depth_mode =
-        regs.depth_mode == Tegra::Engines::Maxwell3D::Regs::DepthMode::ZeroToOne
-            ? GL_ZERO_TO_ONE
-            : GL_NEGATIVE_ONE_TO_ONE;
+void RasterizerOpenGL::SyncViewport() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    const auto& regs = gpu.regs;
+
+    const bool dirty_viewport = flags[Dirty::Viewports];
+    if (dirty_viewport || flags[Dirty::ClipControl]) {
+        flags[Dirty::ClipControl] = false;
+
+        bool flip_y = false;
+        if (regs.viewport_transform[0].scale_y < 0.0) {
+            flip_y = !flip_y;
+        }
+        if (regs.screen_y_control.y_negate != 0) {
+            flip_y = !flip_y;
+        }
+        glClipControl(flip_y ? GL_UPPER_LEFT : GL_LOWER_LEFT,
+                      regs.depth_mode == Maxwell::DepthMode::ZeroToOne ? GL_ZERO_TO_ONE
+                                                                       : GL_NEGATIVE_ONE_TO_ONE);
+    }
+
+    if (dirty_viewport) {
+        flags[Dirty::Viewports] = false;
+
+        const bool force = flags[Dirty::ViewportTransform];
+        flags[Dirty::ViewportTransform] = false;
+
+        for (std::size_t i = 0; i < Maxwell::NumViewports; ++i) {
+            if (!force && !flags[Dirty::Viewport0 + i]) {
+                continue;
+            }
+            flags[Dirty::Viewport0 + i] = false;
+
+            const auto& src = regs.viewport_transform[i];
+            const Common::Rectangle<f32> rect{src.GetRect()};
+            glViewportIndexedf(static_cast<GLuint>(i), rect.left, rect.bottom, rect.GetWidth(),
+                               rect.GetHeight());
+
+            const GLdouble reduce_z = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
+            const GLdouble near_depth = src.translate_z - src.scale_z * reduce_z;
+            const GLdouble far_depth = src.translate_z + src.scale_z;
+            glDepthRangeIndexed(static_cast<GLuint>(i), near_depth, far_depth);
+        }
+    }
 }
 
-void RasterizerOpenGL::SyncClipEnabled(
-    const std::array<bool, Maxwell::Regs::NumClipDistances>& clip_mask) {
+void RasterizerOpenGL::SyncDepthClamp() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::DepthClampEnabled]) {
+        return;
+    }
+    flags[Dirty::DepthClampEnabled] = false;
 
-    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,
-        regs.clip_distance_enabled.c4 != 0, regs.clip_distance_enabled.c5 != 0,
-        regs.clip_distance_enabled.c6 != 0, regs.clip_distance_enabled.c7 != 0};
+    const auto& state = gpu.regs.view_volume_clip_control;
+    UNIMPLEMENTED_IF_MSG(state.depth_clamp_far != state.depth_clamp_near,
+                         "Unimplemented depth clamp separation!");
+
+    oglEnable(GL_DEPTH_CLAMP, state.depth_clamp_far || state.depth_clamp_near);
+}
+
+void RasterizerOpenGL::SyncClipEnabled(u32 clip_mask) {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::ClipDistances] && !flags[Dirty::Shaders]) {
+        return;
+    }
+    flags[Dirty::ClipDistances] = false;
+
+    clip_mask &= gpu.regs.clip_distance_enabled;
+    if (clip_mask == last_clip_distance_mask) {
+        return;
+    }
+    last_clip_distance_mask = clip_mask;
 
     for (std::size_t i = 0; i < Maxwell::Regs::NumClipDistances; ++i) {
-        state.clip_distance[i] = reg_state[i] && clip_mask[i];
+        oglEnable(static_cast<GLenum>(GL_CLIP_DISTANCE0 + i), (clip_mask >> i) & 1);
     }
 }
 
@@ -1052,247 +985,442 @@ void RasterizerOpenGL::SyncClipCoef() {
 }
 
 void RasterizerOpenGL::SyncCullMode() {
-    const auto& regs = system.GPU().Maxwell3D().regs;
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    const auto& regs = gpu.regs;
 
-    state.cull.enabled = regs.cull.enabled != 0;
-    if (state.cull.enabled) {
-        state.cull.mode = MaxwellToGL::CullFace(regs.cull.cull_face);
+    if (flags[Dirty::CullTest]) {
+        flags[Dirty::CullTest] = false;
+
+        if (regs.cull_test_enabled) {
+            glEnable(GL_CULL_FACE);
+            glCullFace(MaxwellToGL::CullFace(regs.cull_face));
+        } else {
+            glDisable(GL_CULL_FACE);
+        }
     }
 
-    state.cull.front_face = MaxwellToGL::FrontFace(regs.cull.front_face);
+    if (flags[Dirty::FrontFace]) {
+        flags[Dirty::FrontFace] = false;
+        glFrontFace(MaxwellToGL::FrontFace(regs.front_face));
+    }
 }
 
 void RasterizerOpenGL::SyncPrimitiveRestart() {
-    const auto& regs = system.GPU().Maxwell3D().regs;
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::PrimitiveRestart]) {
+        return;
+    }
+    flags[Dirty::PrimitiveRestart] = false;
 
-    state.primitive_restart.enabled = regs.primitive_restart.enabled;
-    state.primitive_restart.index = regs.primitive_restart.index;
+    if (gpu.regs.primitive_restart.enabled) {
+        glEnable(GL_PRIMITIVE_RESTART);
+        glPrimitiveRestartIndex(gpu.regs.primitive_restart.index);
+    } else {
+        glDisable(GL_PRIMITIVE_RESTART);
+    }
 }
 
 void RasterizerOpenGL::SyncDepthTestState() {
-    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;
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
 
-    if (!state.depth.test_enabled) {
-        return;
+    const auto& regs = gpu.regs;
+    if (flags[Dirty::DepthMask]) {
+        flags[Dirty::DepthMask] = false;
+        glDepthMask(regs.depth_write_enabled ? GL_TRUE : GL_FALSE);
     }
 
-    state.depth.test_func = MaxwellToGL::ComparisonOp(regs.depth_test_func);
+    if (flags[Dirty::DepthTest]) {
+        flags[Dirty::DepthTest] = false;
+        if (regs.depth_test_enable) {
+            glEnable(GL_DEPTH_TEST);
+            glDepthFunc(MaxwellToGL::ComparisonOp(regs.depth_test_func));
+        } else {
+            glDisable(GL_DEPTH_TEST);
+        }
+    }
 }
 
 void RasterizerOpenGL::SyncStencilTestState() {
-    auto& maxwell3d = system.GPU().Maxwell3D();
-    if (!maxwell3d.dirty.stencil_test) {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::StencilTest]) {
         return;
     }
-    maxwell3d.dirty.stencil_test = false;
-
-    const auto& regs = maxwell3d.regs;
-    state.stencil.test_enabled = regs.stencil_enable != 0;
-    state.MarkDirtyStencilState();
+    flags[Dirty::StencilTest] = false;
 
+    const auto& regs = gpu.regs;
     if (!regs.stencil_enable) {
+        glDisable(GL_STENCIL_TEST);
         return;
     }
 
-    state.stencil.front.test_func = MaxwellToGL::ComparisonOp(regs.stencil_front_func_func);
-    state.stencil.front.test_ref = regs.stencil_front_func_ref;
-    state.stencil.front.test_mask = regs.stencil_front_func_mask;
-    state.stencil.front.action_stencil_fail = MaxwellToGL::StencilOp(regs.stencil_front_op_fail);
-    state.stencil.front.action_depth_fail = MaxwellToGL::StencilOp(regs.stencil_front_op_zfail);
-    state.stencil.front.action_depth_pass = MaxwellToGL::StencilOp(regs.stencil_front_op_zpass);
-    state.stencil.front.write_mask = regs.stencil_front_mask;
+    glEnable(GL_STENCIL_TEST);
+    glStencilFuncSeparate(GL_FRONT, MaxwellToGL::ComparisonOp(regs.stencil_front_func_func),
+                          regs.stencil_front_func_ref, regs.stencil_front_func_mask);
+    glStencilOpSeparate(GL_FRONT, MaxwellToGL::StencilOp(regs.stencil_front_op_fail),
+                        MaxwellToGL::StencilOp(regs.stencil_front_op_zfail),
+                        MaxwellToGL::StencilOp(regs.stencil_front_op_zpass));
+    glStencilMaskSeparate(GL_FRONT, regs.stencil_front_mask);
+
     if (regs.stencil_two_side_enable) {
-        state.stencil.back.test_func = MaxwellToGL::ComparisonOp(regs.stencil_back_func_func);
-        state.stencil.back.test_ref = regs.stencil_back_func_ref;
-        state.stencil.back.test_mask = regs.stencil_back_func_mask;
-        state.stencil.back.action_stencil_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_fail);
-        state.stencil.back.action_depth_fail = MaxwellToGL::StencilOp(regs.stencil_back_op_zfail);
-        state.stencil.back.action_depth_pass = MaxwellToGL::StencilOp(regs.stencil_back_op_zpass);
-        state.stencil.back.write_mask = regs.stencil_back_mask;
+        glStencilFuncSeparate(GL_BACK, MaxwellToGL::ComparisonOp(regs.stencil_back_func_func),
+                              regs.stencil_back_func_ref, regs.stencil_back_func_mask);
+        glStencilOpSeparate(GL_BACK, MaxwellToGL::StencilOp(regs.stencil_back_op_fail),
+                            MaxwellToGL::StencilOp(regs.stencil_back_op_zfail),
+                            MaxwellToGL::StencilOp(regs.stencil_back_op_zpass));
+        glStencilMaskSeparate(GL_BACK, regs.stencil_back_mask);
     } else {
-        state.stencil.back.test_func = GL_ALWAYS;
-        state.stencil.back.test_ref = 0;
-        state.stencil.back.test_mask = 0xFFFFFFFF;
-        state.stencil.back.write_mask = 0xFFFFFFFF;
-        state.stencil.back.action_stencil_fail = GL_KEEP;
-        state.stencil.back.action_depth_fail = GL_KEEP;
-        state.stencil.back.action_depth_pass = GL_KEEP;
+        glStencilFuncSeparate(GL_BACK, GL_ALWAYS, 0, 0xFFFFFFFF);
+        glStencilOpSeparate(GL_BACK, GL_KEEP, GL_KEEP, GL_KEEP);
+        glStencilMaskSeparate(GL_BACK, 0xFFFFFFFF);
     }
 }
 
-void RasterizerOpenGL::SyncRasterizeEnable(OpenGLState& current_state) {
-    const auto& regs = system.GPU().Maxwell3D().regs;
-    current_state.rasterizer_discard = regs.rasterize_enable == 0;
+void RasterizerOpenGL::SyncRasterizeEnable() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::RasterizeEnable]) {
+        return;
+    }
+    flags[Dirty::RasterizeEnable] = false;
+
+    oglEnable(GL_RASTERIZER_DISCARD, gpu.regs.rasterize_enable == 0);
+}
+
+void RasterizerOpenGL::SyncPolygonModes() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::PolygonModes]) {
+        return;
+    }
+    flags[Dirty::PolygonModes] = false;
+
+    if (gpu.regs.fill_rectangle) {
+        if (!GLAD_GL_NV_fill_rectangle) {
+            LOG_ERROR(Render_OpenGL, "GL_NV_fill_rectangle used and not supported");
+            glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
+            return;
+        }
+
+        flags[Dirty::PolygonModeFront] = true;
+        flags[Dirty::PolygonModeBack] = true;
+        glPolygonMode(GL_FRONT_AND_BACK, GL_FILL_RECTANGLE_NV);
+        return;
+    }
+
+    if (gpu.regs.polygon_mode_front == gpu.regs.polygon_mode_back) {
+        flags[Dirty::PolygonModeFront] = false;
+        flags[Dirty::PolygonModeBack] = false;
+        glPolygonMode(GL_FRONT_AND_BACK, MaxwellToGL::PolygonMode(gpu.regs.polygon_mode_front));
+        return;
+    }
+
+    if (flags[Dirty::PolygonModeFront]) {
+        flags[Dirty::PolygonModeFront] = false;
+        glPolygonMode(GL_FRONT, MaxwellToGL::PolygonMode(gpu.regs.polygon_mode_front));
+    }
+
+    if (flags[Dirty::PolygonModeBack]) {
+        flags[Dirty::PolygonModeBack] = false;
+        glPolygonMode(GL_BACK, MaxwellToGL::PolygonMode(gpu.regs.polygon_mode_back));
+    }
 }
 
 void RasterizerOpenGL::SyncColorMask() {
-    auto& maxwell3d = system.GPU().Maxwell3D();
-    if (!maxwell3d.dirty.color_mask) {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::ColorMasks]) {
         return;
     }
-    const auto& regs = maxwell3d.regs;
+    flags[Dirty::ColorMasks] = false;
+
+    const bool force = flags[Dirty::ColorMaskCommon];
+    flags[Dirty::ColorMaskCommon] = false;
+
+    const auto& regs = gpu.regs;
+    if (regs.color_mask_common) {
+        if (!force && !flags[Dirty::ColorMask0]) {
+            return;
+        }
+        flags[Dirty::ColorMask0] = false;
 
-    const std::size_t count =
-        regs.independent_blend_enable ? Tegra::Engines::Maxwell3D::Regs::NumRenderTargets : 1;
-    for (std::size_t i = 0; i < count; i++) {
-        const auto& source = regs.color_mask[regs.color_mask_common ? 0 : i];
-        auto& dest = state.color_mask[i];
-        dest.red_enabled = (source.R == 0) ? GL_FALSE : GL_TRUE;
-        dest.green_enabled = (source.G == 0) ? GL_FALSE : GL_TRUE;
-        dest.blue_enabled = (source.B == 0) ? GL_FALSE : GL_TRUE;
-        dest.alpha_enabled = (source.A == 0) ? GL_FALSE : GL_TRUE;
+        auto& mask = regs.color_mask[0];
+        glColorMask(mask.R != 0, mask.B != 0, mask.G != 0, mask.A != 0);
+        return;
     }
 
-    state.MarkDirtyColorMask();
-    maxwell3d.dirty.color_mask = false;
+    // Path without color_mask_common set
+    for (std::size_t i = 0; i < Maxwell::NumRenderTargets; ++i) {
+        if (!force && !flags[Dirty::ColorMask0 + i]) {
+            continue;
+        }
+        flags[Dirty::ColorMask0 + i] = false;
+
+        const auto& mask = regs.color_mask[i];
+        glColorMaski(static_cast<GLuint>(i), mask.R != 0, mask.G != 0, mask.B != 0, mask.A != 0);
+    }
 }
 
 void RasterizerOpenGL::SyncMultiSampleState() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::MultisampleControl]) {
+        return;
+    }
+    flags[Dirty::MultisampleControl] = false;
+
     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;
+    oglEnable(GL_SAMPLE_ALPHA_TO_COVERAGE, regs.multisample_control.alpha_to_coverage);
+    oglEnable(GL_SAMPLE_ALPHA_TO_ONE, regs.multisample_control.alpha_to_one);
 }
 
 void RasterizerOpenGL::SyncFragmentColorClampState() {
-    const auto& regs = system.GPU().Maxwell3D().regs;
-    state.fragment_color_clamp.enabled = regs.frag_color_clamp != 0;
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::FragmentClampColor]) {
+        return;
+    }
+    flags[Dirty::FragmentClampColor] = false;
+
+    glClampColor(GL_CLAMP_FRAGMENT_COLOR, gpu.regs.frag_color_clamp ? GL_TRUE : GL_FALSE);
 }
 
 void RasterizerOpenGL::SyncBlendState() {
-    auto& maxwell3d = system.GPU().Maxwell3D();
-    if (!maxwell3d.dirty.blend_state) {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    const auto& regs = gpu.regs;
+
+    if (flags[Dirty::BlendColor]) {
+        flags[Dirty::BlendColor] = false;
+        glBlendColor(regs.blend_color.r, regs.blend_color.g, regs.blend_color.b,
+                     regs.blend_color.a);
+    }
+
+    // TODO(Rodrigo): Revisit blending, there are several registers we are not reading
+
+    if (!flags[Dirty::BlendStates]) {
         return;
     }
-    const auto& regs = maxwell3d.regs;
-
-    state.blend_color.red = regs.blend_color.r;
-    state.blend_color.green = regs.blend_color.g;
-    state.blend_color.blue = regs.blend_color.b;
-    state.blend_color.alpha = regs.blend_color.a;
-
-    state.independant_blend.enabled = regs.independent_blend_enable;
-    if (!state.independant_blend.enabled) {
-        auto& blend = state.blend[0];
-        const auto& src = regs.blend;
-        blend.enabled = src.enable[0] != 0;
-        if (blend.enabled) {
-            blend.rgb_equation = MaxwellToGL::BlendEquation(src.equation_rgb);
-            blend.src_rgb_func = MaxwellToGL::BlendFunc(src.factor_source_rgb);
-            blend.dst_rgb_func = MaxwellToGL::BlendFunc(src.factor_dest_rgb);
-            blend.a_equation = MaxwellToGL::BlendEquation(src.equation_a);
-            blend.src_a_func = MaxwellToGL::BlendFunc(src.factor_source_a);
-            blend.dst_a_func = MaxwellToGL::BlendFunc(src.factor_dest_a);
-        }
-        for (std::size_t i = 1; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
-            state.blend[i].enabled = false;
+    flags[Dirty::BlendStates] = false;
+
+    if (!regs.independent_blend_enable) {
+        if (!regs.blend.enable[0]) {
+            glDisable(GL_BLEND);
+            return;
         }
-        maxwell3d.dirty.blend_state = false;
-        state.MarkDirtyBlendState();
+        glEnable(GL_BLEND);
+        glBlendFuncSeparate(MaxwellToGL::BlendFunc(regs.blend.factor_source_rgb),
+                            MaxwellToGL::BlendFunc(regs.blend.factor_dest_rgb),
+                            MaxwellToGL::BlendFunc(regs.blend.factor_source_a),
+                            MaxwellToGL::BlendFunc(regs.blend.factor_dest_a));
+        glBlendEquationSeparate(MaxwellToGL::BlendEquation(regs.blend.equation_rgb),
+                                MaxwellToGL::BlendEquation(regs.blend.equation_a));
         return;
     }
 
-    for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
-        auto& blend = state.blend[i];
-        const auto& src = regs.independent_blend[i];
-        blend.enabled = regs.blend.enable[i] != 0;
-        if (!blend.enabled)
+    const bool force = flags[Dirty::BlendIndependentEnabled];
+    flags[Dirty::BlendIndependentEnabled] = false;
+
+    for (std::size_t i = 0; i < Maxwell::NumRenderTargets; ++i) {
+        if (!force && !flags[Dirty::BlendState0 + i]) {
             continue;
-        blend.rgb_equation = MaxwellToGL::BlendEquation(src.equation_rgb);
-        blend.src_rgb_func = MaxwellToGL::BlendFunc(src.factor_source_rgb);
-        blend.dst_rgb_func = MaxwellToGL::BlendFunc(src.factor_dest_rgb);
-        blend.a_equation = MaxwellToGL::BlendEquation(src.equation_a);
-        blend.src_a_func = MaxwellToGL::BlendFunc(src.factor_source_a);
-        blend.dst_a_func = MaxwellToGL::BlendFunc(src.factor_dest_a);
-    }
+        }
+        flags[Dirty::BlendState0 + i] = false;
+
+        if (!regs.blend.enable[i]) {
+            glDisablei(GL_BLEND, static_cast<GLuint>(i));
+            continue;
+        }
+        glEnablei(GL_BLEND, static_cast<GLuint>(i));
 
-    state.MarkDirtyBlendState();
-    maxwell3d.dirty.blend_state = false;
+        const auto& src = regs.independent_blend[i];
+        glBlendFuncSeparatei(static_cast<GLuint>(i), MaxwellToGL::BlendFunc(src.factor_source_rgb),
+                             MaxwellToGL::BlendFunc(src.factor_dest_rgb),
+                             MaxwellToGL::BlendFunc(src.factor_source_a),
+                             MaxwellToGL::BlendFunc(src.factor_dest_a));
+        glBlendEquationSeparatei(static_cast<GLuint>(i),
+                                 MaxwellToGL::BlendEquation(src.equation_rgb),
+                                 MaxwellToGL::BlendEquation(src.equation_a));
+    }
 }
 
 void RasterizerOpenGL::SyncLogicOpState() {
-    const auto& regs = system.GPU().Maxwell3D().regs;
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::LogicOp]) {
+        return;
+    }
+    flags[Dirty::LogicOp] = false;
 
-    state.logic_op.enabled = regs.logic_op.enable != 0;
+    const auto& regs = gpu.regs;
+    if (regs.logic_op.enable) {
+        glEnable(GL_COLOR_LOGIC_OP);
+        glLogicOp(MaxwellToGL::LogicOp(regs.logic_op.operation));
+    } else {
+        glDisable(GL_COLOR_LOGIC_OP);
+    }
+}
 
-    if (!state.logic_op.enabled)
+void RasterizerOpenGL::SyncScissorTest() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::Scissors]) {
         return;
+    }
+    flags[Dirty::Scissors] = false;
 
-    ASSERT_MSG(regs.blend.enable[0] == 0,
-               "Blending and logic op can't be enabled at the same time.");
-
-    state.logic_op.operation = MaxwellToGL::LogicOp(regs.logic_op.operation);
-}
+    const auto& regs = gpu.regs;
+    for (std::size_t index = 0; index < Maxwell::NumViewports; ++index) {
+        if (!flags[Dirty::Scissor0 + index]) {
+            continue;
+        }
+        flags[Dirty::Scissor0 + index] = false;
 
-void RasterizerOpenGL::SyncScissorTest(OpenGLState& current_state) {
-    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 =
-        geometry_shaders_enabled ? Tegra::Engines::Maxwell3D::Regs::NumViewports : 1;
-    for (std::size_t i = 0; i < viewport_count; i++) {
-        const auto& src = regs.scissor_test[i];
-        auto& dst = current_state.viewports[i].scissor;
-        dst.enabled = (src.enable != 0);
-        if (dst.enabled == 0) {
-            return;
+        const auto& src = regs.scissor_test[index];
+        if (src.enable) {
+            glEnablei(GL_SCISSOR_TEST, static_cast<GLuint>(index));
+            glScissorIndexed(static_cast<GLuint>(index), src.min_x, src.min_y,
+                             src.max_x - src.min_x, src.max_y - src.min_y);
+        } else {
+            glDisablei(GL_SCISSOR_TEST, static_cast<GLuint>(index));
         }
-        const u32 width = src.max_x - src.min_x;
-        const u32 height = src.max_y - src.min_y;
-        dst.x = src.min_x;
-        dst.y = src.min_y;
-        dst.width = width;
-        dst.height = height;
     }
 }
 
-void RasterizerOpenGL::SyncTransformFeedback() {
-    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 = system.GPU().Maxwell3D().regs;
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::PointSize]) {
+        return;
+    }
+    flags[Dirty::PointSize] = false;
+
+    oglEnable(GL_POINT_SPRITE, gpu.regs.point_sprite_enable);
+
+    if (gpu.regs.vp_point_size.enable) {
+        // By definition of GL_POINT_SIZE, it only matters if GL_PROGRAM_POINT_SIZE is disabled.
+        glEnable(GL_PROGRAM_POINT_SIZE);
+        return;
+    }
+
     // Limit the point size to 1 since nouveau sometimes sets a point size of 0 (and that's invalid
     // in OpenGL).
-    state.point.program_control = regs.vp_point_size.enable != 0;
-    state.point.sprite = regs.point_sprite_enable != 0;
-    state.point.size = std::max(1.0f, regs.point_size);
+    glPointSize(std::max(1.0f, gpu.regs.point_size));
+    glDisable(GL_PROGRAM_POINT_SIZE);
 }
 
 void RasterizerOpenGL::SyncPolygonOffset() {
-    auto& maxwell3d = system.GPU().Maxwell3D();
-    if (!maxwell3d.dirty.polygon_offset) {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::PolygonOffset]) {
         return;
     }
-    const auto& regs = 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;
+    flags[Dirty::PolygonOffset] = false;
 
-    // Hardware divides polygon offset units by two
-    state.polygon_offset.units = regs.polygon_offset_units / 2.0f;
-    state.polygon_offset.factor = regs.polygon_offset_factor;
-    state.polygon_offset.clamp = regs.polygon_offset_clamp;
+    const auto& regs = gpu.regs;
+    oglEnable(GL_POLYGON_OFFSET_FILL, regs.polygon_offset_fill_enable);
+    oglEnable(GL_POLYGON_OFFSET_LINE, regs.polygon_offset_line_enable);
+    oglEnable(GL_POLYGON_OFFSET_POINT, regs.polygon_offset_point_enable);
 
-    state.MarkDirtyPolygonOffset();
-    maxwell3d.dirty.polygon_offset = false;
+    if (regs.polygon_offset_fill_enable || regs.polygon_offset_line_enable ||
+        regs.polygon_offset_point_enable) {
+        // Hardware divides polygon offset units by two
+        glPolygonOffsetClamp(regs.polygon_offset_factor, regs.polygon_offset_units / 2.0f,
+                             regs.polygon_offset_clamp);
+    }
 }
 
 void RasterizerOpenGL::SyncAlphaTest() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::AlphaTest]) {
+        return;
+    }
+    flags[Dirty::AlphaTest] = false;
+
+    const auto& regs = gpu.regs;
+    if (regs.alpha_test_enabled && regs.rt_control.count > 1) {
+        LOG_WARNING(Render_OpenGL, "Alpha testing with more than one render target is not tested");
+    }
+
+    if (regs.alpha_test_enabled) {
+        glEnable(GL_ALPHA_TEST);
+        glAlphaFunc(MaxwellToGL::ComparisonOp(regs.alpha_test_func), regs.alpha_test_ref);
+    } else {
+        glDisable(GL_ALPHA_TEST);
+    }
+}
+
+void RasterizerOpenGL::SyncFramebufferSRGB() {
+    auto& gpu = system.GPU().Maxwell3D();
+    auto& flags = gpu.dirty.flags;
+    if (!flags[Dirty::FramebufferSRGB]) {
+        return;
+    }
+    flags[Dirty::FramebufferSRGB] = false;
+
+    oglEnable(GL_FRAMEBUFFER_SRGB, gpu.regs.framebuffer_srgb);
+}
+
+void RasterizerOpenGL::BeginTransformFeedback(GLenum primitive_mode) {
     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");
+    if (regs.tfb_enabled == 0) {
+        return;
+    }
+
+    UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::TesselationControl) ||
+                     regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::TesselationEval) ||
+                     regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::Geometry));
 
-    state.alpha_test.enabled = regs.alpha_test_enabled;
-    if (!state.alpha_test.enabled) {
+    for (std::size_t index = 0; index < Maxwell::NumTransformFeedbackBuffers; ++index) {
+        const auto& binding = regs.tfb_bindings[index];
+        if (!binding.buffer_enable) {
+            if (enabled_transform_feedback_buffers[index]) {
+                glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, static_cast<GLuint>(index), 0, 0,
+                                  0);
+            }
+            enabled_transform_feedback_buffers[index] = false;
+            continue;
+        }
+        enabled_transform_feedback_buffers[index] = true;
+
+        auto& tfb_buffer = transform_feedback_buffers[index];
+        tfb_buffer.Create();
+
+        const GLuint handle = tfb_buffer.handle;
+        const std::size_t size = binding.buffer_size;
+        glNamedBufferData(handle, static_cast<GLsizeiptr>(size), nullptr, GL_STREAM_COPY);
+        glBindBufferRange(GL_TRANSFORM_FEEDBACK_BUFFER, static_cast<GLuint>(index), handle, 0,
+                          static_cast<GLsizeiptr>(size));
+    }
+
+    glBeginTransformFeedback(GL_POINTS);
+}
+
+void RasterizerOpenGL::EndTransformFeedback() {
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    if (regs.tfb_enabled == 0) {
         return;
     }
-    state.alpha_test.func = MaxwellToGL::ComparisonOp(regs.alpha_test_func);
-    state.alpha_test.ref = regs.alpha_test_ref;
+
+    glEndTransformFeedback();
+
+    for (std::size_t index = 0; index < Maxwell::NumTransformFeedbackBuffers; ++index) {
+        const auto& binding = regs.tfb_bindings[index];
+        if (!binding.buffer_enable) {
+            continue;
+        }
+        UNIMPLEMENTED_IF(binding.buffer_offset != 0);
+
+        const GLuint handle = transform_feedback_buffers[index].handle;
+        const GPUVAddr gpu_addr = binding.Address();
+        const std::size_t size = binding.buffer_size;
+        const auto [dest_buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
+        glCopyNamedBufferSubData(handle, *dest_buffer, 0, offset, static_cast<GLsizeiptr>(size));
+    }
 }
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 68abe9a21..2d3be2437 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -30,7 +30,7 @@
 #include "video_core/renderer_opengl/gl_shader_cache.h"
 #include "video_core/renderer_opengl/gl_shader_decompiler.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/textures/texture.h"
@@ -55,7 +55,8 @@ struct DrawParameters;
 class RasterizerOpenGL : public VideoCore::RasterizerAccelerated {
 public:
     explicit RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
-                              ScreenInfo& info);
+                              ScreenInfo& info, GLShader::ProgramManager& program_manager,
+                              StateTracker& state_tracker);
     ~RasterizerOpenGL() override;
 
     void Draw(bool is_indexed, bool is_instanced) override;
@@ -76,6 +77,7 @@ public:
                            u32 pixel_stride) override;
     void LoadDiskResources(const std::atomic_bool& stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;
+    void SetupDirtyFlags() override;
 
     /// Returns true when there are commands queued to the OpenGL server.
     bool AnyCommandQueued() const {
@@ -86,8 +88,7 @@ private:
     /// Configures the color and depth framebuffer states.
     void ConfigureFramebuffers();
 
-    void ConfigureClearFramebuffer(OpenGLState& current_state, bool using_color_fb,
-                                   bool using_depth_fb, bool using_stencil_fb);
+    void ConfigureClearFramebuffer(bool using_color_fb, bool using_depth_fb, bool using_stencil_fb);
 
     /// Configures the current constbuffers to use for the draw command.
     void SetupDrawConstBuffers(std::size_t stage_index, const Shader& shader);
@@ -97,7 +98,7 @@ private:
 
     /// Configures a constant buffer.
     void SetupConstBuffer(u32 binding, const Tegra::Engines::ConstBufferInfo& buffer,
-                          const GLShader::ConstBufferEntry& entry);
+                          const ConstBufferEntry& entry);
 
     /// Configures the current global memory entries to use for the draw command.
     void SetupDrawGlobalMemory(std::size_t stage_index, const Shader& shader);
@@ -106,7 +107,7 @@ private:
     void SetupComputeGlobalMemory(const Shader& kernel);
 
     /// Configures a constant buffer.
-    void SetupGlobalMemory(u32 binding, const GLShader::GlobalMemoryEntry& entry, GPUVAddr gpu_addr,
+    void SetupGlobalMemory(u32 binding, const GlobalMemoryEntry& entry, GPUVAddr gpu_addr,
                            std::size_t size);
 
     /// Configures the current textures to use for the draw command.
@@ -117,7 +118,7 @@ private:
 
     /// Configures a texture.
     void SetupTexture(u32 binding, const Tegra::Texture::FullTextureInfo& texture,
-                      const GLShader::SamplerEntry& entry);
+                      const SamplerEntry& entry);
 
     /// Configures images in a graphics shader.
     void SetupDrawImages(std::size_t stage_index, const Shader& shader);
@@ -126,15 +127,16 @@ private:
     void SetupComputeImages(const Shader& shader);
 
     /// Configures an image.
-    void SetupImage(u32 binding, const Tegra::Texture::TICEntry& tic,
-                    const GLShader::ImageEntry& entry);
+    void SetupImage(u32 binding, const Tegra::Texture::TICEntry& tic, const ImageEntry& entry);
 
     /// Syncs the viewport and depth range to match the guest state
-    void SyncViewport(OpenGLState& current_state);
+    void SyncViewport();
+
+    /// Syncs the depth clamp state
+    void SyncDepthClamp();
 
     /// Syncs the clip enabled status to match the guest state
-    void SyncClipEnabled(
-        const std::array<bool, Tegra::Engines::Maxwell3D::Regs::NumClipDistances>& clip_mask);
+    void SyncClipEnabled(u32 clip_mask);
 
     /// Syncs the clip coefficients to match the guest state
     void SyncClipCoef();
@@ -164,16 +166,16 @@ private:
     void SyncMultiSampleState();
 
     /// Syncs the scissor test state to match the guest state
-    void SyncScissorTest(OpenGLState& current_state);
-
-    /// Syncs the transform feedback state to match the guest state
-    void SyncTransformFeedback();
+    void SyncScissorTest();
 
     /// Syncs the point state to match the guest state
     void SyncPointState();
 
     /// Syncs the rasterizer enable state to match the guest state
-    void SyncRasterizeEnable(OpenGLState& current_state);
+    void SyncRasterizeEnable();
+
+    /// Syncs polygon modes to match the guest state
+    void SyncPolygonModes();
 
     /// Syncs Color Mask
     void SyncColorMask();
@@ -184,6 +186,15 @@ private:
     /// Syncs the alpha test state to match the guest state
     void SyncAlphaTest();
 
+    /// Syncs the framebuffer sRGB state to match the guest state
+    void SyncFramebufferSRGB();
+
+    /// Begin a transform feedback
+    void BeginTransformFeedback(GLenum primitive_mode);
+
+    /// End a transform feedback
+    void EndTransformFeedback();
+
     /// Check for extension that are not strictly required but are needed for correct emulation
     void CheckExtensions();
 
@@ -191,18 +202,17 @@ private:
 
     std::size_t CalculateIndexBufferSize() const;
 
-    /// Updates and returns a vertex array object representing current vertex format
-    GLuint SetupVertexFormat();
+    /// Updates the current vertex format
+    void SetupVertexFormat();
 
-    void SetupVertexBuffer(GLuint vao);
-    void SetupVertexInstances(GLuint vao);
+    void SetupVertexBuffer();
+    void SetupVertexInstances();
 
     GLintptr SetupIndexBuffer();
 
     void SetupShaders(GLenum primitive_mode);
 
     const Device device;
-    OpenGLState state;
 
     TextureCacheOpenGL texture_cache;
     ShaderCacheOpenGL shader_cache;
@@ -212,22 +222,25 @@ private:
 
     Core::System& system;
     ScreenInfo& screen_info;
-
-    std::unique_ptr<GLShader::ProgramManager> shader_program_manager;
-    std::map<std::array<Tegra::Engines::Maxwell3D::Regs::VertexAttribute,
-                        Tegra::Engines::Maxwell3D::Regs::NumVertexAttributes>,
-             OGLVertexArray>
-        vertex_array_cache;
+    GLShader::ProgramManager& program_manager;
+    StateTracker& state_tracker;
 
     static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
     OGLBufferCache buffer_cache;
 
-    VertexArrayPushBuffer vertex_array_pushbuffer;
+    VertexArrayPushBuffer vertex_array_pushbuffer{state_tracker};
     BindBuffersRangePushBuffer bind_ubo_pushbuffer{GL_UNIFORM_BUFFER};
     BindBuffersRangePushBuffer bind_ssbo_pushbuffer{GL_SHADER_STORAGE_BUFFER};
 
+    std::array<OGLBuffer, Tegra::Engines::Maxwell3D::Regs::NumTransformFeedbackBuffers>
+        transform_feedback_buffers;
+    std::bitset<Tegra::Engines::Maxwell3D::Regs::NumTransformFeedbackBuffers>
+        enabled_transform_feedback_buffers;
+
     /// Number of commands queued to the OpenGL driver. Reseted on flush.
     std::size_t num_queued_commands = 0;
+
+    u32 last_clip_distance_mask = 0;
 };
 
 } // 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 f0ddfb276..97803d480 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp
@@ -8,13 +8,29 @@
 #include "common/microprofile.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_shader_util.h"
-#include "video_core/renderer_opengl/gl_state.h"
 
 MICROPROFILE_DEFINE(OpenGL_ResourceCreation, "OpenGL", "Resource Creation", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_RGB(128, 128, 192));
 
 namespace OpenGL {
 
+void OGLRenderbuffer::Create() {
+    if (handle != 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
+    glCreateRenderbuffers(1, &handle);
+}
+
+void OGLRenderbuffer::Release() {
+    if (handle == 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
+    glDeleteRenderbuffers(1, &handle);
+    handle = 0;
+}
+
 void OGLTexture::Create(GLenum target) {
     if (handle != 0)
         return;
@@ -29,7 +45,6 @@ void OGLTexture::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteTextures(1, &handle);
-    OpenGLState::GetCurState().UnbindTexture(handle).Apply();
     handle = 0;
 }
 
@@ -47,7 +62,6 @@ void OGLTextureView::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteTextures(1, &handle);
-    OpenGLState::GetCurState().UnbindTexture(handle).Apply();
     handle = 0;
 }
 
@@ -65,7 +79,6 @@ void OGLSampler::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteSamplers(1, &handle);
-    OpenGLState::GetCurState().ResetSampler(handle).Apply();
     handle = 0;
 }
 
@@ -109,7 +122,6 @@ void OGLProgram::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteProgram(handle);
-    OpenGLState::GetCurState().ResetProgram(handle).Apply();
     handle = 0;
 }
 
@@ -127,7 +139,6 @@ void OGLPipeline::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteProgramPipelines(1, &handle);
-    OpenGLState::GetCurState().ResetPipeline(handle).Apply();
     handle = 0;
 }
 
@@ -171,24 +182,6 @@ void OGLSync::Release() {
     handle = 0;
 }
 
-void OGLVertexArray::Create() {
-    if (handle != 0)
-        return;
-
-    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
-    glCreateVertexArrays(1, &handle);
-}
-
-void OGLVertexArray::Release() {
-    if (handle == 0)
-        return;
-
-    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
-    glDeleteVertexArrays(1, &handle);
-    OpenGLState::GetCurState().ResetVertexArray(handle).Apply();
-    handle = 0;
-}
-
 void OGLFramebuffer::Create() {
     if (handle != 0)
         return;
@@ -203,7 +196,6 @@ void OGLFramebuffer::Release() {
 
     MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
     glDeleteFramebuffers(1, &handle);
-    OpenGLState::GetCurState().ResetFramebuffer(handle).Apply();
     handle = 0;
 }
 
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h
index 514d1d165..de93f4212 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.h
+++ b/src/video_core/renderer_opengl/gl_resource_manager.h
@@ -11,6 +11,31 @@
 
 namespace OpenGL {
 
+class OGLRenderbuffer : private NonCopyable {
+public:
+    OGLRenderbuffer() = default;
+
+    OGLRenderbuffer(OGLRenderbuffer&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+
+    ~OGLRenderbuffer() {
+        Release();
+    }
+
+    OGLRenderbuffer& operator=(OGLRenderbuffer&& o) noexcept {
+        Release();
+        handle = std::exchange(o.handle, 0);
+        return *this;
+    }
+
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create();
+
+    /// Deletes the internal OpenGL resource
+    void Release();
+
+    GLuint handle = 0;
+};
+
 class OGLTexture : private NonCopyable {
 public:
     OGLTexture() = default;
@@ -216,31 +241,6 @@ public:
     GLsync handle = 0;
 };
 
-class OGLVertexArray : private NonCopyable {
-public:
-    OGLVertexArray() = default;
-
-    OGLVertexArray(OGLVertexArray&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
-
-    ~OGLVertexArray() {
-        Release();
-    }
-
-    OGLVertexArray& operator=(OGLVertexArray&& o) noexcept {
-        Release();
-        handle = std::exchange(o.handle, 0);
-        return *this;
-    }
-
-    /// Creates a new internal OpenGL resource and stores the handle
-    void Create();
-
-    /// Deletes the internal OpenGL resource
-    void Release();
-
-    GLuint handle = 0;
-};
-
 class OGLFramebuffer : private NonCopyable {
 public:
     OGLFramebuffer() = default;
diff --git a/src/video_core/renderer_opengl/gl_sampler_cache.cpp b/src/video_core/renderer_opengl/gl_sampler_cache.cpp
index 3ded5ecea..5c174879a 100644
--- a/src/video_core/renderer_opengl/gl_sampler_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_sampler_cache.cpp
@@ -38,7 +38,7 @@ OGLSampler SamplerCacheOpenGL::CreateSampler(const Tegra::Texture::TSCEntry& tsc
         glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY, tsc.GetMaxAnisotropy());
     } else if (GLAD_GL_EXT_texture_filter_anisotropic) {
         glSamplerParameterf(sampler_id, GL_TEXTURE_MAX_ANISOTROPY_EXT, tsc.GetMaxAnisotropy());
-    } else if (tsc.GetMaxAnisotropy() != 1) {
+    } else {
         LOG_WARNING(Render_OpenGL, "Anisotropy not supported by host GPU driver");
     }
 
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 489eb143c..e3d31c3eb 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -2,12 +2,16 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <atomic>
+#include <functional>
 #include <mutex>
 #include <optional>
 #include <string>
 #include <thread>
 #include <unordered_set>
+
 #include <boost/functional/hash.hpp>
+
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
@@ -22,14 +26,16 @@
 #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_state_tracker.h"
 #include "video_core/renderer_opengl/utils.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace OpenGL {
 
 using Tegra::Engines::ShaderType;
-using VideoCommon::Shader::ConstBufferLocker;
 using VideoCommon::Shader::ProgramCode;
+using VideoCommon::Shader::Registry;
 using VideoCommon::Shader::ShaderIR;
 
 namespace {
@@ -55,7 +61,7 @@ constexpr bool IsSchedInstruction(std::size_t offset, std::size_t main_offset) {
 }
 
 /// Calculates the size of a program stream
-std::size_t CalculateProgramSize(const GLShader::ProgramCode& program) {
+std::size_t CalculateProgramSize(const ProgramCode& program) {
     constexpr std::size_t start_offset = 10;
     // This is the encoded version of BRA that jumps to itself. All Nvidia
     // shaders end with one.
@@ -108,32 +114,9 @@ constexpr GLenum GetGLShaderType(ShaderType shader_type) {
     }
 }
 
-/// Describes primitive behavior on geometry shaders
-constexpr std::pair<const char*, u32> GetPrimitiveDescription(GLenum primitive_mode) {
-    switch (primitive_mode) {
-    case GL_POINTS:
-        return {"points", 1};
-    case GL_LINES:
-    case GL_LINE_STRIP:
-        return {"lines", 2};
-    case GL_LINES_ADJACENCY:
-    case GL_LINE_STRIP_ADJACENCY:
-        return {"lines_adjacency", 4};
-    case GL_TRIANGLES:
-    case GL_TRIANGLE_STRIP:
-    case GL_TRIANGLE_FAN:
-        return {"triangles", 3};
-    case GL_TRIANGLES_ADJACENCY:
-    case GL_TRIANGLE_STRIP_ADJACENCY:
-        return {"triangles_adjacency", 6};
-    default:
-        return {"points", 1};
-    }
-}
-
 /// Hashes one (or two) program streams
 u64 GetUniqueIdentifier(ShaderType shader_type, bool is_a, const ProgramCode& code,
-                        const ProgramCode& code_b) {
+                        const ProgramCode& code_b = {}) {
     u64 unique_identifier = boost::hash_value(code);
     if (is_a) {
         // VertexA programs include two programs
@@ -142,24 +125,6 @@ u64 GetUniqueIdentifier(ShaderType shader_type, bool is_a, const ProgramCode& co
     return unique_identifier;
 }
 
-/// Creates an unspecialized program from code streams
-std::string GenerateGLSL(const Device& device, ShaderType shader_type, const ShaderIR& ir,
-                         const std::optional<ShaderIR>& ir_b) {
-    switch (shader_type) {
-    case ShaderType::Vertex:
-        return GLShader::GenerateVertexShader(device, ir, ir_b ? &*ir_b : nullptr);
-    case ShaderType::Geometry:
-        return GLShader::GenerateGeometryShader(device, ir);
-    case ShaderType::Fragment:
-        return GLShader::GenerateFragmentShader(device, ir);
-    case ShaderType::Compute:
-        return GLShader::GenerateComputeShader(device, ir);
-    default:
-        UNIMPLEMENTED_MSG("Unimplemented shader_type={}", static_cast<u32>(shader_type));
-        return {};
-    }
-}
-
 constexpr const char* GetShaderTypeName(ShaderType shader_type) {
     switch (shader_type) {
     case ShaderType::Vertex:
@@ -195,102 +160,38 @@ constexpr ShaderType GetShaderType(Maxwell::ShaderProgram program_type) {
     return {};
 }
 
-std::string GetShaderId(u64 unique_identifier, ShaderType shader_type) {
+std::string MakeShaderID(u64 unique_identifier, ShaderType shader_type) {
     return fmt::format("{}{:016X}", GetShaderTypeName(shader_type), unique_identifier);
 }
 
-Tegra::Engines::ConstBufferEngineInterface& GetConstBufferEngineInterface(Core::System& system,
-                                                                          ShaderType shader_type) {
-    if (shader_type == ShaderType::Compute) {
-        return system.GPU().KeplerCompute();
-    } else {
-        return system.GPU().Maxwell3D();
-    }
-}
-
-std::unique_ptr<ConstBufferLocker> MakeLocker(Core::System& system, ShaderType shader_type) {
-    return std::make_unique<ConstBufferLocker>(shader_type,
-                                               GetConstBufferEngineInterface(system, shader_type));
-}
-
-void FillLocker(ConstBufferLocker& locker, const ShaderDiskCacheUsage& usage) {
-    locker.SetBoundBuffer(usage.bound_buffer);
-    for (const auto& key : usage.keys) {
-        const auto [buffer, offset] = key.first;
-        locker.InsertKey(buffer, offset, key.second);
+std::shared_ptr<Registry> MakeRegistry(const ShaderDiskCacheEntry& entry) {
+    const VideoCore::GuestDriverProfile guest_profile{entry.texture_handler_size};
+    const VideoCommon::Shader::SerializedRegistryInfo info{guest_profile, entry.bound_buffer,
+                                                           entry.graphics_info, entry.compute_info};
+    const auto registry = std::make_shared<Registry>(entry.type, info);
+    for (const auto& [address, value] : entry.keys) {
+        const auto [buffer, offset] = address;
+        registry->InsertKey(buffer, offset, value);
     }
-    for (const auto& [offset, sampler] : usage.bound_samplers) {
-        locker.InsertBoundSampler(offset, sampler);
+    for (const auto& [offset, sampler] : entry.bound_samplers) {
+        registry->InsertBoundSampler(offset, sampler);
     }
-    for (const auto& [key, sampler] : usage.bindless_samplers) {
+    for (const auto& [key, sampler] : entry.bindless_samplers) {
         const auto [buffer, offset] = key;
-        locker.InsertBindlessSampler(buffer, offset, sampler);
+        registry->InsertBindlessSampler(buffer, offset, sampler);
     }
+    return registry;
 }
 
-CachedProgram BuildShader(const Device& device, u64 unique_identifier, ShaderType shader_type,
-                          const ProgramCode& code, const ProgramCode& code_b,
-                          ConstBufferLocker& locker, const ProgramVariant& variant,
-                          bool hint_retrievable = false) {
-    LOG_INFO(Render_OpenGL, "called. {}", GetShaderId(unique_identifier, shader_type));
-
-    const bool is_compute = shader_type == ShaderType::Compute;
-    const u32 main_offset = is_compute ? KERNEL_MAIN_OFFSET : STAGE_MAIN_OFFSET;
-    const ShaderIR ir(code, main_offset, COMPILER_SETTINGS, locker);
-    std::optional<ShaderIR> ir_b;
-    if (!code_b.empty()) {
-        ir_b.emplace(code_b, main_offset, COMPILER_SETTINGS, locker);
-    }
-
-    std::string source = fmt::format(R"(// {}
-#version 430 core
-#extension GL_ARB_separate_shader_objects : enable
-)",
-                                     GetShaderId(unique_identifier, shader_type));
-    if (device.HasShaderBallot()) {
-        source += "#extension GL_ARB_shader_ballot : require\n";
-    }
-    if (device.HasVertexViewportLayer()) {
-        source += "#extension GL_ARB_shader_viewport_layer_array : require\n";
-    }
-    if (device.HasImageLoadFormatted()) {
-        source += "#extension GL_EXT_shader_image_load_formatted : require\n";
-    }
-    if (device.HasWarpIntrinsics()) {
-        source += "#extension GL_NV_gpu_shader5 : require\n"
-                  "#extension GL_NV_shader_thread_group : require\n"
-                  "#extension GL_NV_shader_thread_shuffle : require\n";
-    }
-    // This pragma stops Nvidia's driver from over optimizing math (probably using fp16 operations)
-    // on places where we don't want to.
-    // Thanks to Ryujinx for finding this workaround.
-    source += "#pragma optionNV(fastmath off)\n";
-
-    if (shader_type == ShaderType::Geometry) {
-        const auto [glsl_topology, max_vertices] = GetPrimitiveDescription(variant.primitive_mode);
-        source += fmt::format("#define MAX_VERTEX_INPUT {}\n", max_vertices);
-        source += fmt::format("layout ({}) in;\n", glsl_topology);
-    }
-    if (shader_type == ShaderType::Compute) {
-        if (variant.local_memory_size > 0) {
-            source += fmt::format("#define LOCAL_MEMORY_SIZE {}\n",
-                                  Common::AlignUp(variant.local_memory_size, 4) / 4);
-        }
-        source +=
-            fmt::format("layout (local_size_x = {}, local_size_y = {}, local_size_z = {}) in;\n",
-                        variant.block_x, variant.block_y, variant.block_z);
-
-        if (variant.shared_memory_size > 0) {
-            // shared_memory_size is described in number of words
-            source += fmt::format("shared uint smem[{}];\n", variant.shared_memory_size);
-        }
-    }
-
-    source += '\n';
-    source += GenerateGLSL(device, shader_type, ir, ir_b);
+std::shared_ptr<OGLProgram> BuildShader(const Device& device, ShaderType shader_type,
+                                        u64 unique_identifier, const ShaderIR& ir,
+                                        const Registry& registry, bool hint_retrievable = false) {
+    const std::string shader_id = MakeShaderID(unique_identifier, shader_type);
+    LOG_INFO(Render_OpenGL, "{}", shader_id);
 
+    const std::string glsl = DecompileShader(device, ir, registry, shader_type, shader_id);
     OGLShader shader;
-    shader.Create(source.c_str(), GetGLShaderType(shader_type));
+    shader.Create(glsl.c_str(), GetGLShaderType(shader_type));
 
     auto program = std::make_shared<OGLProgram>();
     program->Create(true, hint_retrievable, shader.handle);
@@ -298,7 +199,7 @@ CachedProgram BuildShader(const Device& device, u64 unique_identifier, ShaderTyp
 }
 
 std::unordered_set<GLenum> GetSupportedFormats() {
-    GLint num_formats{};
+    GLint num_formats;
     glGetIntegerv(GL_NUM_PROGRAM_BINARY_FORMATS, &num_formats);
 
     std::vector<GLint> formats(num_formats);
@@ -313,115 +214,82 @@ std::unordered_set<GLenum> GetSupportedFormats() {
 
 } // Anonymous namespace
 
-CachedShader::CachedShader(const ShaderParameters& params, ShaderType shader_type,
-                           GLShader::ShaderEntries entries, ProgramCode code, ProgramCode code_b)
-    : RasterizerCacheObject{params.host_ptr}, system{params.system},
-      disk_cache{params.disk_cache}, device{params.device}, cpu_addr{params.cpu_addr},
-      unique_identifier{params.unique_identifier}, shader_type{shader_type},
-      entries{std::move(entries)}, code{std::move(code)}, code_b{std::move(code_b)} {
-    if (!params.precompiled_variants) {
-        return;
-    }
-    for (const auto& pair : *params.precompiled_variants) {
-        auto locker = MakeLocker(system, shader_type);
-        const auto& usage = pair->first;
-        FillLocker(*locker, usage);
-
-        std::unique_ptr<LockerVariant>* locker_variant = nullptr;
-        const auto it =
-            std::find_if(locker_variants.begin(), locker_variants.end(), [&](const auto& variant) {
-                return variant->locker->HasEqualKeys(*locker);
-            });
-        if (it == locker_variants.end()) {
-            locker_variant = &locker_variants.emplace_back();
-            *locker_variant = std::make_unique<LockerVariant>();
-            locker_variant->get()->locker = std::move(locker);
-        } else {
-            locker_variant = &*it;
-        }
-        locker_variant->get()->programs.emplace(usage.variant, pair->second);
-    }
+CachedShader::CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes,
+                           std::shared_ptr<VideoCommon::Shader::Registry> registry,
+                           ShaderEntries entries, std::shared_ptr<OGLProgram> program)
+    : RasterizerCacheObject{host_ptr}, registry{std::move(registry)}, entries{std::move(entries)},
+      cpu_addr{cpu_addr}, size_in_bytes{size_in_bytes}, program{std::move(program)} {}
+
+CachedShader::~CachedShader() = default;
+
+GLuint CachedShader::GetHandle() const {
+    DEBUG_ASSERT(registry->IsConsistent());
+    return program->handle;
 }
 
 Shader CachedShader::CreateStageFromMemory(const ShaderParameters& params,
                                            Maxwell::ShaderProgram program_type, ProgramCode code,
                                            ProgramCode code_b) {
     const auto shader_type = GetShaderType(program_type);
-    params.disk_cache.SaveRaw(
-        ShaderDiskCacheRaw(params.unique_identifier, shader_type, code, code_b));
+    const std::size_t size_in_bytes = code.size() * sizeof(u64);
 
-    ConstBufferLocker locker(shader_type, params.system.GPU().Maxwell3D());
-    const ShaderIR ir(code, STAGE_MAIN_OFFSET, COMPILER_SETTINGS, locker);
+    auto registry = std::make_shared<Registry>(shader_type, params.system.GPU().Maxwell3D());
+    const ShaderIR ir(code, STAGE_MAIN_OFFSET, COMPILER_SETTINGS, *registry);
     // TODO(Rodrigo): Handle VertexA shaders
     // std::optional<ShaderIR> ir_b;
     // if (!code_b.empty()) {
     //     ir_b.emplace(code_b, STAGE_MAIN_OFFSET);
     // }
-    return std::shared_ptr<CachedShader>(new CachedShader(
-        params, shader_type, GLShader::GetEntries(ir), std::move(code), std::move(code_b)));
+    auto program = BuildShader(params.device, shader_type, params.unique_identifier, ir, *registry);
+
+    ShaderDiskCacheEntry entry;
+    entry.type = shader_type;
+    entry.code = std::move(code);
+    entry.code_b = std::move(code_b);
+    entry.unique_identifier = params.unique_identifier;
+    entry.bound_buffer = registry->GetBoundBuffer();
+    entry.graphics_info = registry->GetGraphicsInfo();
+    entry.keys = registry->GetKeys();
+    entry.bound_samplers = registry->GetBoundSamplers();
+    entry.bindless_samplers = registry->GetBindlessSamplers();
+    params.disk_cache.SaveEntry(std::move(entry));
+
+    return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr,
+                                                          size_in_bytes, std::move(registry),
+                                                          MakeEntries(ir), std::move(program)));
 }
 
 Shader CachedShader::CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code) {
-    params.disk_cache.SaveRaw(
-        ShaderDiskCacheRaw(params.unique_identifier, ShaderType::Compute, code));
-
-    ConstBufferLocker locker(Tegra::Engines::ShaderType::Compute,
-                             params.system.GPU().KeplerCompute());
-    const ShaderIR ir(code, KERNEL_MAIN_OFFSET, COMPILER_SETTINGS, locker);
-    return std::shared_ptr<CachedShader>(new CachedShader(
-        params, ShaderType::Compute, GLShader::GetEntries(ir), std::move(code), {}));
+    const std::size_t size_in_bytes = code.size() * sizeof(u64);
+
+    auto& engine = params.system.GPU().KeplerCompute();
+    auto registry = std::make_shared<Registry>(ShaderType::Compute, engine);
+    const ShaderIR ir(code, KERNEL_MAIN_OFFSET, COMPILER_SETTINGS, *registry);
+    const u64 uid = params.unique_identifier;
+    auto program = BuildShader(params.device, ShaderType::Compute, uid, ir, *registry);
+
+    ShaderDiskCacheEntry entry;
+    entry.type = ShaderType::Compute;
+    entry.code = std::move(code);
+    entry.unique_identifier = uid;
+    entry.bound_buffer = registry->GetBoundBuffer();
+    entry.compute_info = registry->GetComputeInfo();
+    entry.keys = registry->GetKeys();
+    entry.bound_samplers = registry->GetBoundSamplers();
+    entry.bindless_samplers = registry->GetBindlessSamplers();
+    params.disk_cache.SaveEntry(std::move(entry));
+
+    return std::shared_ptr<CachedShader>(new CachedShader(params.host_ptr, params.cpu_addr,
+                                                          size_in_bytes, std::move(registry),
+                                                          MakeEntries(ir), std::move(program)));
 }
 
 Shader CachedShader::CreateFromCache(const ShaderParameters& params,
-                                     const UnspecializedShader& unspecialized) {
-    return std::shared_ptr<CachedShader>(new CachedShader(params, unspecialized.type,
-                                                          unspecialized.entries, unspecialized.code,
-                                                          unspecialized.code_b));
-}
-
-GLuint CachedShader::GetHandle(const ProgramVariant& variant) {
-    EnsureValidLockerVariant();
-
-    const auto [entry, is_cache_miss] = curr_locker_variant->programs.try_emplace(variant);
-    auto& program = entry->second;
-    if (!is_cache_miss) {
-        return program->handle;
-    }
-
-    program = BuildShader(device, unique_identifier, shader_type, code, code_b,
-                          *curr_locker_variant->locker, variant);
-    disk_cache.SaveUsage(GetUsage(variant, *curr_locker_variant->locker));
-
-    LabelGLObject(GL_PROGRAM, program->handle, cpu_addr);
-    return program->handle;
-}
-
-bool CachedShader::EnsureValidLockerVariant() {
-    const auto previous_variant = curr_locker_variant;
-    if (curr_locker_variant && !curr_locker_variant->locker->IsConsistent()) {
-        curr_locker_variant = nullptr;
-    }
-    if (!curr_locker_variant) {
-        for (auto& variant : locker_variants) {
-            if (variant->locker->IsConsistent()) {
-                curr_locker_variant = variant.get();
-            }
-        }
-    }
-    if (!curr_locker_variant) {
-        auto& new_variant = locker_variants.emplace_back();
-        new_variant = std::make_unique<LockerVariant>();
-        new_variant->locker = MakeLocker(system, shader_type);
-        curr_locker_variant = new_variant.get();
-    }
-    return previous_variant == curr_locker_variant;
-}
-
-ShaderDiskCacheUsage CachedShader::GetUsage(const ProgramVariant& variant,
-                                            const ConstBufferLocker& locker) const {
-    return ShaderDiskCacheUsage{unique_identifier,         variant,
-                                locker.GetBoundBuffer(),   locker.GetKeys(),
-                                locker.GetBoundSamplers(), locker.GetBindlessSamplers()};
+                                     const PrecompiledShader& precompiled_shader,
+                                     std::size_t size_in_bytes) {
+    return std::shared_ptr<CachedShader>(new CachedShader(
+        params.host_ptr, params.cpu_addr, size_in_bytes, precompiled_shader.registry,
+        precompiled_shader.entries, precompiled_shader.program));
 }
 
 ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
@@ -431,16 +299,12 @@ ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System&
 
 void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
                                       const VideoCore::DiskResourceLoadCallback& callback) {
-    const auto transferable = disk_cache.LoadTransferable();
+    const std::optional transferable = disk_cache.LoadTransferable();
     if (!transferable) {
         return;
     }
-    const auto [raws, shader_usages] = *transferable;
-    if (!GenerateUnspecializedShaders(stop_loading, callback, raws) || stop_loading) {
-        return;
-    }
 
-    const auto dumps = disk_cache.LoadPrecompiled();
+    const std::vector gl_cache = disk_cache.LoadPrecompiled();
     const auto supported_formats = GetSupportedFormats();
 
     // Track if precompiled cache was altered during loading to know if we have to
@@ -449,77 +313,82 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
 
     // Inform the frontend about shader build initialization
     if (callback) {
-        callback(VideoCore::LoadCallbackStage::Build, 0, shader_usages.size());
+        callback(VideoCore::LoadCallbackStage::Build, 0, transferable->size());
     }
 
     std::mutex mutex;
     std::size_t built_shaders = 0; // It doesn't have be atomic since it's used behind a mutex
-    std::atomic_bool compilation_failed = false;
+    std::atomic_bool gl_cache_failed = false;
 
-    const auto Worker = [&](Core::Frontend::GraphicsContext* context, std::size_t begin,
-                            std::size_t end, const std::vector<ShaderDiskCacheUsage>& shader_usages,
-                            const ShaderDumpsMap& dumps) {
+    const auto find_precompiled = [&gl_cache](u64 id) {
+        return std::find_if(gl_cache.begin(), gl_cache.end(),
+                            [id](const auto& entry) { return entry.unique_identifier == id; });
+    };
+
+    const auto worker = [&](Core::Frontend::GraphicsContext* context, std::size_t begin,
+                            std::size_t end) {
         context->MakeCurrent();
         SCOPE_EXIT({ return context->DoneCurrent(); });
 
         for (std::size_t i = begin; i < end; ++i) {
-            if (stop_loading || compilation_failed) {
+            if (stop_loading) {
                 return;
             }
-            const auto& usage{shader_usages[i]};
-            const auto& unspecialized{unspecialized_shaders.at(usage.unique_identifier)};
-            const auto dump{dumps.find(usage)};
-
-            CachedProgram shader;
-            if (dump != dumps.end()) {
-                // If the shader is dumped, attempt to load it with
-                shader = GeneratePrecompiledProgram(dump->second, supported_formats);
-                if (!shader) {
-                    compilation_failed = true;
-                    return;
+            const auto& entry = (*transferable)[i];
+            const u64 uid = entry.unique_identifier;
+            const auto it = find_precompiled(uid);
+            const auto precompiled_entry = it != gl_cache.end() ? &*it : nullptr;
+
+            const bool is_compute = entry.type == ShaderType::Compute;
+            const u32 main_offset = is_compute ? KERNEL_MAIN_OFFSET : STAGE_MAIN_OFFSET;
+            auto registry = MakeRegistry(entry);
+            const ShaderIR ir(entry.code, main_offset, COMPILER_SETTINGS, *registry);
+
+            std::shared_ptr<OGLProgram> program;
+            if (precompiled_entry) {
+                // If the shader is precompiled, attempt to load it with
+                program = GeneratePrecompiledProgram(entry, *precompiled_entry, supported_formats);
+                if (!program) {
+                    gl_cache_failed = true;
                 }
             }
-            if (!shader) {
-                auto locker{MakeLocker(system, unspecialized.type)};
-                FillLocker(*locker, usage);
-
-                shader = BuildShader(device, usage.unique_identifier, unspecialized.type,
-                                     unspecialized.code, unspecialized.code_b, *locker,
-                                     usage.variant, true);
+            if (!program) {
+                // Otherwise compile it from GLSL
+                program = BuildShader(device, entry.type, uid, ir, *registry, true);
             }
 
+            PrecompiledShader shader;
+            shader.program = std::move(program);
+            shader.registry = std::move(registry);
+            shader.entries = MakeEntries(ir);
+
             std::scoped_lock lock{mutex};
             if (callback) {
                 callback(VideoCore::LoadCallbackStage::Build, ++built_shaders,
-                         shader_usages.size());
+                         transferable->size());
             }
-
-            precompiled_programs.emplace(usage, std::move(shader));
-
-            // TODO(Rodrigo): Is there a better way to do this?
-            precompiled_variants[usage.unique_identifier].push_back(
-                precompiled_programs.find(usage));
+            runtime_cache.emplace(entry.unique_identifier, std::move(shader));
         }
     };
 
     const auto num_workers{static_cast<std::size_t>(std::thread::hardware_concurrency() + 1ULL)};
-    const std::size_t bucket_size{shader_usages.size() / num_workers};
+    const std::size_t bucket_size{transferable->size() / num_workers};
     std::vector<std::unique_ptr<Core::Frontend::GraphicsContext>> contexts(num_workers);
     std::vector<std::thread> threads(num_workers);
     for (std::size_t i = 0; i < num_workers; ++i) {
         const bool is_last_worker = i + 1 == num_workers;
         const std::size_t start{bucket_size * i};
-        const std::size_t end{is_last_worker ? shader_usages.size() : start + bucket_size};
+        const std::size_t end{is_last_worker ? transferable->size() : start + bucket_size};
 
         // On some platforms the shared context has to be created from the GUI thread
         contexts[i] = emu_window.CreateSharedContext();
-        threads[i] = std::thread(Worker, contexts[i].get(), start, end, shader_usages, dumps);
+        threads[i] = std::thread(worker, contexts[i].get(), start, end);
     }
     for (auto& thread : threads) {
         thread.join();
     }
 
-    if (compilation_failed) {
+    if (gl_cache_failed) {
         // Invalidate the precompiled cache if a shader dumped shader was rejected
         disk_cache.InvalidatePrecompiled();
         precompiled_cache_altered = true;
@@ -532,11 +401,12 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
     // TODO(Rodrigo): Do state tracking for transferable shaders and do a dummy draw
     // before precompiling them
 
-    for (std::size_t i = 0; i < shader_usages.size(); ++i) {
-        const auto& usage{shader_usages[i]};
-        if (dumps.find(usage) == dumps.end()) {
-            const auto& program{precompiled_programs.at(usage)};
-            disk_cache.SaveDump(usage, program->handle);
+    for (std::size_t i = 0; i < transferable->size(); ++i) {
+        const u64 id = (*transferable)[i].unique_identifier;
+        const auto it = find_precompiled(id);
+        if (it == gl_cache.end()) {
+            const GLuint program = runtime_cache.at(id).program->handle;
+            disk_cache.SavePrecompiled(id, program);
             precompiled_cache_altered = true;
         }
     }
@@ -546,84 +416,33 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
     }
 }
 
-const PrecompiledVariants* ShaderCacheOpenGL::GetPrecompiledVariants(u64 unique_identifier) const {
-    const auto it = precompiled_variants.find(unique_identifier);
-    return it == precompiled_variants.end() ? nullptr : &it->second;
-}
-
-CachedProgram ShaderCacheOpenGL::GeneratePrecompiledProgram(
-    const ShaderDiskCacheDump& dump, const std::unordered_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");
+std::shared_ptr<OGLProgram> ShaderCacheOpenGL::GeneratePrecompiledProgram(
+    const ShaderDiskCacheEntry& entry, const ShaderDiskCachePrecompiled& precompiled_entry,
+    const std::unordered_set<GLenum>& supported_formats) {
+    if (supported_formats.find(precompiled_entry.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);
+    auto program = std::make_shared<OGLProgram>();
+    program->handle = glCreateProgram();
+    glProgramParameteri(program->handle, GL_PROGRAM_SEPARABLE, GL_TRUE);
+    glProgramBinary(program->handle, precompiled_entry.binary_format,
+                    precompiled_entry.binary.data(),
+                    static_cast<GLsizei>(precompiled_entry.binary.size()));
+
+    GLint link_status;
+    glGetProgramiv(program->handle, GL_LINK_STATUS, &link_status);
     if (link_status == GL_FALSE) {
-        LOG_INFO(Render_OpenGL, "Precompiled cache rejected by the driver - removing");
+        LOG_INFO(Render_OpenGL, "Precompiled cache rejected by the driver, removing");
         return {};
     }
 
-    return shader;
-}
-
-bool ShaderCacheOpenGL::GenerateUnspecializedShaders(
-    const std::atomic_bool& stop_loading, const VideoCore::DiskResourceLoadCallback& callback,
-    const std::vector<ShaderDiskCacheRaw>& raws) {
-    if (callback) {
-        callback(VideoCore::LoadCallbackStage::Decompile, 0, raws.size());
-    }
-
-    for (std::size_t i = 0; i < raws.size(); ++i) {
-        if (stop_loading) {
-            return false;
-        }
-        const auto& raw{raws[i]};
-        const u64 unique_identifier{raw.GetUniqueIdentifier()};
-        const u64 calculated_hash{
-            GetUniqueIdentifier(raw.GetType(), raw.HasProgramA(), raw.GetCode(), raw.GetCodeB())};
-        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 false;
-        }
-
-        const u32 main_offset =
-            raw.GetType() == ShaderType::Compute ? KERNEL_MAIN_OFFSET : STAGE_MAIN_OFFSET;
-        ConstBufferLocker locker(raw.GetType());
-        const ShaderIR ir(raw.GetCode(), main_offset, COMPILER_SETTINGS, locker);
-        // TODO(Rodrigo): Handle VertexA shaders
-        // std::optional<ShaderIR> ir_b;
-        // if (raw.HasProgramA()) {
-        //     ir_b.emplace(raw.GetProgramCodeB(), main_offset);
-        // }
-
-        UnspecializedShader unspecialized;
-        unspecialized.entries = GLShader::GetEntries(ir);
-        unspecialized.type = raw.GetType();
-        unspecialized.code = raw.GetCode();
-        unspecialized.code_b = raw.GetCodeB();
-        unspecialized_shaders.emplace(raw.GetUniqueIdentifier(), unspecialized);
-
-        if (callback) {
-            callback(VideoCore::LoadCallbackStage::Decompile, i, raws.size());
-        }
-    }
-    return true;
+    return program;
 }
 
 Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
-    if (!system.GPU().Maxwell3D().dirty.shaders) {
+    if (!system.GPU().Maxwell3D().dirty.flags[Dirty::Shaders]) {
         return last_shaders[static_cast<std::size_t>(program)];
     }
 
@@ -647,17 +466,17 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
 
     const auto unique_identifier = GetUniqueIdentifier(
         GetShaderType(program), program == Maxwell::ShaderProgram::VertexA, code, code_b);
-    const auto precompiled_variants = GetPrecompiledVariants(unique_identifier);
     const auto cpu_addr{*memory_manager.GpuToCpuAddress(address)};
-    const ShaderParameters params{system,   disk_cache, precompiled_variants, device,
+    const ShaderParameters params{system,   disk_cache, device,
                                   cpu_addr, host_ptr,   unique_identifier};
 
-    const auto found = unspecialized_shaders.find(unique_identifier);
-    if (found == unspecialized_shaders.end()) {
+    const auto found = runtime_cache.find(unique_identifier);
+    if (found == runtime_cache.end()) {
         shader = CachedShader::CreateStageFromMemory(params, program, std::move(code),
                                                      std::move(code_b));
     } else {
-        shader = CachedShader::CreateFromCache(params, found->second);
+        const std::size_t size_in_bytes = code.size() * sizeof(u64);
+        shader = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
     }
     Register(shader);
 
@@ -672,19 +491,19 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
         return kernel;
     }
 
-    // No kernel found - create a new one
+    // No kernel found, create a new one
     auto code{GetShaderCode(memory_manager, code_addr, host_ptr)};
-    const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code, {})};
-    const auto precompiled_variants = GetPrecompiledVariants(unique_identifier);
+    const auto unique_identifier{GetUniqueIdentifier(ShaderType::Compute, false, code)};
     const auto cpu_addr{*memory_manager.GpuToCpuAddress(code_addr)};
-    const ShaderParameters params{system,   disk_cache, precompiled_variants, device,
+    const ShaderParameters params{system,   disk_cache, device,
                                   cpu_addr, host_ptr,   unique_identifier};
 
-    const auto found = unspecialized_shaders.find(unique_identifier);
-    if (found == unspecialized_shaders.end()) {
+    const auto found = runtime_cache.find(unique_identifier);
+    if (found == runtime_cache.end()) {
         kernel = CachedShader::CreateKernelFromMemory(params, std::move(code));
     } else {
-        kernel = CachedShader::CreateFromCache(params, found->second);
+        const std::size_t size_in_bytes = code.size() * sizeof(u64);
+        kernel = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
     }
 
     Register(kernel);
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index 7b1470db3..4935019fc 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -22,7 +22,7 @@
 #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/shader/const_buffer_locker.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace Core {
@@ -41,22 +41,17 @@ class RasterizerOpenGL;
 struct UnspecializedShader;
 
 using Shader = std::shared_ptr<CachedShader>;
-using CachedProgram = std::shared_ptr<OGLProgram>;
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-using PrecompiledPrograms = std::unordered_map<ShaderDiskCacheUsage, CachedProgram>;
-using PrecompiledVariants = std::vector<PrecompiledPrograms::iterator>;
-
-struct UnspecializedShader {
-    GLShader::ShaderEntries entries;
-    Tegra::Engines::ShaderType type;
-    ProgramCode code;
-    ProgramCode code_b;
+
+struct PrecompiledShader {
+    std::shared_ptr<OGLProgram> program;
+    std::shared_ptr<VideoCommon::Shader::Registry> registry;
+    ShaderEntries entries;
 };
 
 struct ShaderParameters {
     Core::System& system;
     ShaderDiskCacheOpenGL& disk_cache;
-    const PrecompiledVariants* precompiled_variants;
     const Device& device;
     VAddr cpu_addr;
     u8* host_ptr;
@@ -65,61 +60,45 @@ struct ShaderParameters {
 
 class CachedShader final : public RasterizerCacheObject {
 public:
-    static Shader CreateStageFromMemory(const ShaderParameters& params,
-                                        Maxwell::ShaderProgram program_type,
-                                        ProgramCode program_code, ProgramCode program_code_b);
-    static Shader CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code);
+    ~CachedShader();
 
-    static Shader CreateFromCache(const ShaderParameters& params,
-                                  const UnspecializedShader& unspecialized);
+    /// Gets the GL program handle for the shader
+    GLuint GetHandle() const;
 
+    /// Returns the guest CPU address of the shader
     VAddr GetCpuAddr() const override {
         return cpu_addr;
     }
 
+    /// Returns the size in bytes of the shader
     std::size_t GetSizeInBytes() const override {
-        return code.size() * sizeof(u64);
+        return size_in_bytes;
     }
 
     /// Gets the shader entries for the shader
-    const GLShader::ShaderEntries& GetShaderEntries() const {
+    const ShaderEntries& GetEntries() const {
         return entries;
     }
 
-    /// Gets the GL program handle for the shader
-    GLuint GetHandle(const ProgramVariant& variant);
-
-private:
-    struct LockerVariant {
-        std::unique_ptr<VideoCommon::Shader::ConstBufferLocker> locker;
-        std::unordered_map<ProgramVariant, CachedProgram> programs;
-    };
-
-    explicit CachedShader(const ShaderParameters& params, Tegra::Engines::ShaderType shader_type,
-                          GLShader::ShaderEntries entries, ProgramCode program_code,
-                          ProgramCode program_code_b);
-
-    bool EnsureValidLockerVariant();
-
-    ShaderDiskCacheUsage GetUsage(const ProgramVariant& variant,
-                                  const VideoCommon::Shader::ConstBufferLocker& locker) const;
-
-    Core::System& system;
-    ShaderDiskCacheOpenGL& disk_cache;
-    const Device& device;
-
-    VAddr cpu_addr{};
-
-    u64 unique_identifier{};
-    Tegra::Engines::ShaderType shader_type{};
-
-    GLShader::ShaderEntries entries;
+    static Shader CreateStageFromMemory(const ShaderParameters& params,
+                                        Maxwell::ShaderProgram program_type,
+                                        ProgramCode program_code, ProgramCode program_code_b);
+    static Shader CreateKernelFromMemory(const ShaderParameters& params, ProgramCode code);
 
-    ProgramCode code;
-    ProgramCode code_b;
+    static Shader CreateFromCache(const ShaderParameters& params,
+                                  const PrecompiledShader& precompiled_shader,
+                                  std::size_t size_in_bytes);
 
-    LockerVariant* curr_locker_variant = nullptr;
-    std::vector<std::unique_ptr<LockerVariant>> locker_variants;
+private:
+    explicit CachedShader(const u8* host_ptr, VAddr cpu_addr, std::size_t size_in_bytes,
+                          std::shared_ptr<VideoCommon::Shader::Registry> registry,
+                          ShaderEntries entries, std::shared_ptr<OGLProgram> program);
+
+    std::shared_ptr<VideoCommon::Shader::Registry> registry;
+    ShaderEntries entries;
+    VAddr cpu_addr = 0;
+    std::size_t size_in_bytes = 0;
+    std::shared_ptr<OGLProgram> program;
 };
 
 class ShaderCacheOpenGL final : public RasterizerCache<Shader> {
@@ -142,25 +121,15 @@ protected:
     void FlushObjectInner(const Shader& object) override {}
 
 private:
-    bool GenerateUnspecializedShaders(const std::atomic_bool& stop_loading,
-                                      const VideoCore::DiskResourceLoadCallback& callback,
-                                      const std::vector<ShaderDiskCacheRaw>& raws);
-
-    CachedProgram GeneratePrecompiledProgram(const ShaderDiskCacheDump& dump,
-                                             const std::unordered_set<GLenum>& supported_formats);
-
-    const PrecompiledVariants* GetPrecompiledVariants(u64 unique_identifier) const;
+    std::shared_ptr<OGLProgram> GeneratePrecompiledProgram(
+        const ShaderDiskCacheEntry& entry, const ShaderDiskCachePrecompiled& precompiled_entry,
+        const std::unordered_set<GLenum>& supported_formats);
 
     Core::System& system;
     Core::Frontend::EmuWindow& emu_window;
     const Device& device;
-
     ShaderDiskCacheOpenGL disk_cache;
-
-    PrecompiledPrograms precompiled_programs;
-    std::unordered_map<u64, PrecompiledVariants> precompiled_variants;
-
-    std::unordered_map<u64, UnspecializedShader> unspecialized_shaders;
+    std::unordered_map<u64, PrecompiledShader> runtime_cache;
 
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 };
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 4735000b5..8aa4a7ac9 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -23,8 +23,9 @@
 #include "video_core/shader/ast.h"
 #include "video_core/shader/node.h"
 #include "video_core/shader/shader_ir.h"
+#include "video_core/shader/transform_feedback.h"
 
-namespace OpenGL::GLShader {
+namespace OpenGL {
 
 namespace {
 
@@ -36,6 +37,8 @@ using Tegra::Shader::IpaInterpMode;
 using Tegra::Shader::IpaMode;
 using Tegra::Shader::IpaSampleMode;
 using Tegra::Shader::Register;
+using VideoCommon::Shader::BuildTransformFeedback;
+using VideoCommon::Shader::Registry;
 
 using namespace std::string_literals;
 using namespace VideoCommon::Shader;
@@ -48,6 +51,11 @@ class ExprDecompiler;
 
 enum class Type { Void, Bool, Bool2, Float, Int, Uint, HalfFloat };
 
+constexpr std::array FLOAT_TYPES{"float", "vec2", "vec3", "vec4"};
+
+constexpr std::string_view INPUT_ATTRIBUTE_NAME = "in_attr";
+constexpr std::string_view OUTPUT_ATTRIBUTE_NAME = "out_attr";
+
 struct TextureOffset {};
 struct TextureDerivates {};
 using TextureArgument = std::pair<Type, Node>;
@@ -56,6 +64,25 @@ using TextureIR = std::variant<TextureOffset, TextureDerivates, TextureArgument>
 constexpr u32 MAX_CONSTBUFFER_ELEMENTS =
     static_cast<u32>(Maxwell::MaxConstBufferSize) / (4 * sizeof(float));
 
+constexpr std::string_view CommonDeclarations = R"(#define ftoi floatBitsToInt
+#define ftou floatBitsToUint
+#define itof intBitsToFloat
+#define utof uintBitsToFloat
+
+bvec2 HalfFloatNanComparison(bvec2 comparison, vec2 pair1, vec2 pair2) {{
+    bvec2 is_nan1 = isnan(pair1);
+    bvec2 is_nan2 = isnan(pair2);
+    return bvec2(comparison.x || is_nan1.x || is_nan2.x, comparison.y || is_nan1.y || is_nan2.y);
+}}
+
+const float fswzadd_modifiers_a[] = float[4](-1.0f,  1.0f, -1.0f,  0.0f );
+const float fswzadd_modifiers_b[] = float[4](-1.0f, -1.0f,  1.0f, -1.0f );
+
+layout (std140, binding = {}) uniform vs_config {{
+    float y_direction;
+}};
+)";
+
 class ShaderWriter final {
 public:
     void AddExpression(std::string_view text) {
@@ -269,12 +296,41 @@ const char* GetImageTypeDeclaration(Tegra::Shader::ImageType image_type) {
     }
 }
 
+/// Describes primitive behavior on geometry shaders
+std::pair<const char*, u32> GetPrimitiveDescription(Maxwell::PrimitiveTopology topology) {
+    switch (topology) {
+    case Maxwell::PrimitiveTopology::Points:
+        return {"points", 1};
+    case Maxwell::PrimitiveTopology::Lines:
+    case Maxwell::PrimitiveTopology::LineStrip:
+        return {"lines", 2};
+    case Maxwell::PrimitiveTopology::LinesAdjacency:
+    case Maxwell::PrimitiveTopology::LineStripAdjacency:
+        return {"lines_adjacency", 4};
+    case Maxwell::PrimitiveTopology::Triangles:
+    case Maxwell::PrimitiveTopology::TriangleStrip:
+    case Maxwell::PrimitiveTopology::TriangleFan:
+        return {"triangles", 3};
+    case Maxwell::PrimitiveTopology::TrianglesAdjacency:
+    case Maxwell::PrimitiveTopology::TriangleStripAdjacency:
+        return {"triangles_adjacency", 6};
+    default:
+        UNIMPLEMENTED_MSG("topology={}", static_cast<int>(topology));
+        return {"points", 1};
+    }
+}
+
 /// Generates code to use for a swizzle operation.
-constexpr const char* GetSwizzle(u32 element) {
+constexpr const char* GetSwizzle(std::size_t element) {
     constexpr std::array swizzle = {".x", ".y", ".z", ".w"};
     return swizzle.at(element);
 }
 
+constexpr const char* GetColorSwizzle(std::size_t element) {
+    constexpr std::array swizzle = {".r", ".g", ".b", ".a"};
+    return swizzle.at(element);
+}
+
 /// Translate topology
 std::string GetTopologyName(Tegra::Shader::OutputTopology topology) {
     switch (topology) {
@@ -310,10 +366,19 @@ constexpr bool IsGenericAttribute(Attribute::Index index) {
     return index >= Attribute::Index::Attribute_0 && index <= Attribute::Index::Attribute_31;
 }
 
+constexpr bool IsLegacyTexCoord(Attribute::Index index) {
+    return static_cast<int>(index) >= static_cast<int>(Attribute::Index::TexCoord_0) &&
+           static_cast<int>(index) <= static_cast<int>(Attribute::Index::TexCoord_7);
+}
+
 constexpr Attribute::Index ToGenericAttribute(u64 value) {
     return static_cast<Attribute::Index>(value + static_cast<u64>(Attribute::Index::Attribute_0));
 }
 
+constexpr int GetLegacyTexCoordIndex(Attribute::Index index) {
+    return static_cast<int>(index) - static_cast<int>(Attribute::Index::TexCoord_0);
+}
+
 u32 GetGenericAttributeIndex(Attribute::Index index) {
     ASSERT(IsGenericAttribute(index));
     return static_cast<u32>(index) - static_cast<u32>(Attribute::Index::Attribute_0);
@@ -337,15 +402,66 @@ std::string FlowStackTopName(MetaStackClass stack) {
     return fmt::format("{}_flow_stack_top", GetFlowStackPrefix(stack));
 }
 
-[[deprecated]] constexpr bool IsVertexShader(ShaderType stage) {
-    return stage == ShaderType::Vertex;
-}
+struct GenericVaryingDescription {
+    std::string name;
+    u8 first_element = 0;
+    bool is_scalar = false;
+};
 
 class GLSLDecompiler final {
 public:
-    explicit GLSLDecompiler(const Device& device, const ShaderIR& ir, ShaderType stage,
-                            std::string suffix)
-        : device{device}, ir{ir}, stage{stage}, suffix{suffix}, header{ir.GetHeader()} {}
+    explicit GLSLDecompiler(const Device& device, const ShaderIR& ir, const Registry& registry,
+                            ShaderType stage, std::string_view identifier, std::string_view suffix)
+        : device{device}, ir{ir}, registry{registry}, stage{stage},
+          identifier{identifier}, suffix{suffix}, header{ir.GetHeader()} {
+        if (stage != ShaderType::Compute) {
+            transform_feedback = BuildTransformFeedback(registry.GetGraphicsInfo());
+        }
+    }
+
+    void Decompile() {
+        DeclareHeader();
+        DeclareVertex();
+        DeclareGeometry();
+        DeclareFragment();
+        DeclareCompute();
+        DeclareInputAttributes();
+        DeclareOutputAttributes();
+        DeclareImages();
+        DeclareSamplers();
+        DeclareGlobalMemory();
+        DeclareConstantBuffers();
+        DeclareLocalMemory();
+        DeclareRegisters();
+        DeclarePredicates();
+        DeclareInternalFlags();
+        DeclareCustomVariables();
+        DeclarePhysicalAttributeReader();
+
+        code.AddLine("void main() {{");
+        ++code.scope;
+
+        if (stage == ShaderType::Vertex) {
+            code.AddLine("gl_Position = vec4(0.0f, 0.0f, 0.0f, 1.0f);");
+        }
+
+        if (ir.IsDecompiled()) {
+            DecompileAST();
+        } else {
+            DecompileBranchMode();
+        }
+
+        --code.scope;
+        code.AddLine("}}");
+    }
+
+    std::string GetResult() {
+        return code.GetResult();
+    }
+
+private:
+    friend class ASTDecompiler;
+    friend class ExprDecompiler;
 
     void DecompileBranchMode() {
         // VM's program counter
@@ -387,46 +503,40 @@ public:
 
     void DecompileAST();
 
-    void Decompile() {
-        DeclareVertex();
-        DeclareGeometry();
-        DeclareRegisters();
-        DeclareCustomVariables();
-        DeclarePredicates();
-        DeclareLocalMemory();
-        DeclareInternalFlags();
-        DeclareInputAttributes();
-        DeclareOutputAttributes();
-        DeclareConstantBuffers();
-        DeclareGlobalMemory();
-        DeclareSamplers();
-        DeclareImages();
-        DeclarePhysicalAttributeReader();
-
-        code.AddLine("void execute_{}() {{", suffix);
-        ++code.scope;
-
-        if (ir.IsDecompiled()) {
-            DecompileAST();
-        } else {
-            DecompileBranchMode();
+    void DeclareHeader() {
+        if (!identifier.empty()) {
+            code.AddLine("// {}", identifier);
+        }
+        code.AddLine("#version 440 {}", ir.UsesLegacyVaryings() ? "compatibility" : "core");
+        code.AddLine("#extension GL_ARB_separate_shader_objects : enable");
+        if (device.HasShaderBallot()) {
+            code.AddLine("#extension GL_ARB_shader_ballot : require");
+        }
+        if (device.HasVertexViewportLayer()) {
+            code.AddLine("#extension GL_ARB_shader_viewport_layer_array : require");
+        }
+        if (device.HasImageLoadFormatted()) {
+            code.AddLine("#extension GL_EXT_shader_image_load_formatted : require");
         }
+        if (device.HasWarpIntrinsics()) {
+            code.AddLine("#extension GL_NV_gpu_shader5 : require");
+            code.AddLine("#extension GL_NV_shader_thread_group : require");
+            code.AddLine("#extension GL_NV_shader_thread_shuffle : require");
+        }
+        // This pragma stops Nvidia's driver from over optimizing math (probably using fp16
+        // operations) on places where we don't want to.
+        // Thanks to Ryujinx for finding this workaround.
+        code.AddLine("#pragma optionNV(fastmath off)");
 
-        --code.scope;
-        code.AddLine("}}");
-    }
+        code.AddNewLine();
 
-    std::string GetResult() {
-        return code.GetResult();
+        code.AddLine(CommonDeclarations, EmulationUniformBlockBinding);
     }
 
-private:
-    friend class ASTDecompiler;
-    friend class ExprDecompiler;
-
     void DeclareVertex() {
-        if (!IsVertexShader(stage))
+        if (stage != ShaderType::Vertex) {
             return;
+        }
 
         DeclareVertexRedeclarations();
     }
@@ -436,9 +546,15 @@ private:
             return;
         }
 
+        const auto& info = registry.GetGraphicsInfo();
+        const auto input_topology = info.primitive_topology;
+        const auto [glsl_topology, max_vertices] = GetPrimitiveDescription(input_topology);
+        max_input_vertices = max_vertices;
+        code.AddLine("layout ({}) in;", glsl_topology);
+
         const auto topology = GetTopologyName(header.common3.output_topology);
-        const auto max_vertices = header.common4.max_output_vertices.Value();
-        code.AddLine("layout ({}, max_vertices = {}) out;", topology, max_vertices);
+        const auto max_output_vertices = header.common4.max_output_vertices.Value();
+        code.AddLine("layout ({}, max_vertices = {}) out;", topology, max_output_vertices);
         code.AddNewLine();
 
         code.AddLine("in gl_PerVertex {{");
@@ -450,11 +566,50 @@ private:
         DeclareVertexRedeclarations();
     }
 
+    void DeclareFragment() {
+        if (stage != ShaderType::Fragment) {
+            return;
+        }
+        if (ir.UsesLegacyVaryings()) {
+            code.AddLine("in gl_PerFragment {{");
+            ++code.scope;
+            code.AddLine("vec4 gl_TexCoord[8];");
+            code.AddLine("vec4 gl_Color;");
+            code.AddLine("vec4 gl_SecondaryColor;");
+            --code.scope;
+            code.AddLine("}};");
+        }
+
+        for (u32 rt = 0; rt < Maxwell::NumRenderTargets; ++rt) {
+            code.AddLine("layout (location = {}) out vec4 frag_color{};", rt, rt);
+        }
+    }
+
+    void DeclareCompute() {
+        if (stage != ShaderType::Compute) {
+            return;
+        }
+        const auto& info = registry.GetComputeInfo();
+        if (const u32 size = info.shared_memory_size_in_words; size > 0) {
+            code.AddLine("shared uint smem[{}];", size);
+            code.AddNewLine();
+        }
+        code.AddLine("layout (local_size_x = {}, local_size_y = {}, local_size_z = {}) in;",
+                     info.workgroup_size[0], info.workgroup_size[1], info.workgroup_size[2]);
+        code.AddNewLine();
+    }
+
     void DeclareVertexRedeclarations() {
         code.AddLine("out gl_PerVertex {{");
         ++code.scope;
 
-        code.AddLine("vec4 gl_Position;");
+        auto pos_xfb = GetTransformFeedbackDecoration(Attribute::Index::Position);
+        if (!pos_xfb.empty()) {
+            pos_xfb = fmt::format("layout ({}) ", pos_xfb);
+        }
+        const char* pos_type =
+            FLOAT_TYPES.at(GetNumComponents(Attribute::Index::Position).value_or(4) - 1);
+        code.AddLine("{}{} gl_Position;", pos_xfb, pos_type);
 
         for (const auto attribute : ir.GetOutputAttributes()) {
             if (attribute == Attribute::Index::ClipDistances0123 ||
@@ -463,14 +618,14 @@ private:
                 break;
             }
         }
-        if (!IsVertexShader(stage) || device.HasVertexViewportLayer()) {
+        if (stage != ShaderType::Vertex || device.HasVertexViewportLayer()) {
             if (ir.UsesLayer()) {
                 code.AddLine("int gl_Layer;");
             }
             if (ir.UsesViewportIndex()) {
                 code.AddLine("int gl_ViewportIndex;");
             }
-        } else if ((ir.UsesLayer() || ir.UsesViewportIndex()) && IsVertexShader(stage) &&
+        } else if ((ir.UsesLayer() || ir.UsesViewportIndex()) && stage == ShaderType::Vertex &&
                    !device.HasVertexViewportLayer()) {
             LOG_ERROR(
                 Render_OpenGL,
@@ -481,12 +636,12 @@ private:
             code.AddLine("float gl_PointSize;");
         }
 
-        if (ir.UsesInstanceId()) {
-            code.AddLine("int gl_InstanceID;");
-        }
-
-        if (ir.UsesVertexId()) {
-            code.AddLine("int gl_VertexID;");
+        if (ir.UsesLegacyVaryings()) {
+            code.AddLine("vec4 gl_TexCoord[8];");
+            code.AddLine("vec4 gl_FrontColor;");
+            code.AddLine("vec4 gl_FrontSecondaryColor;");
+            code.AddLine("vec4 gl_BackColor;");
+            code.AddLine("vec4 gl_BackSecondaryColor;");
         }
 
         --code.scope;
@@ -525,18 +680,16 @@ private:
     }
 
     void DeclareLocalMemory() {
+        u64 local_memory_size = 0;
         if (stage == ShaderType::Compute) {
-            code.AddLine("#ifdef LOCAL_MEMORY_SIZE");
-            code.AddLine("uint {}[LOCAL_MEMORY_SIZE];", GetLocalMemory());
-            code.AddLine("#endif");
-            return;
+            local_memory_size = registry.GetComputeInfo().local_memory_size_in_words * 4ULL;
+        } else {
+            local_memory_size = header.GetLocalMemorySize();
         }
-
-        const u64 local_memory_size = header.GetLocalMemorySize();
         if (local_memory_size == 0) {
             return;
         }
-        const auto element_count = Common::AlignUp(local_memory_size, 4) / 4;
+        const u64 element_count = Common::AlignUp(local_memory_size, 4) / 4;
         code.AddLine("uint {}[{}];", GetLocalMemory(), element_count);
         code.AddNewLine();
     }
@@ -589,7 +742,7 @@ private:
     void DeclareInputAttribute(Attribute::Index index, bool skip_unused) {
         const u32 location{GetGenericAttributeIndex(index)};
 
-        std::string name{GetInputAttribute(index)};
+        std::string name{GetGenericInputAttribute(index)};
         if (stage == ShaderType::Geometry) {
             name = "gs_" + name + "[]";
         }
@@ -626,9 +779,59 @@ private:
         }
     }
 
+    std::optional<std::size_t> GetNumComponents(Attribute::Index index, u8 element = 0) const {
+        const u8 location = static_cast<u8>(static_cast<u32>(index) * 4 + element);
+        const auto it = transform_feedback.find(location);
+        if (it == transform_feedback.end()) {
+            return {};
+        }
+        return it->second.components;
+    }
+
+    std::string GetTransformFeedbackDecoration(Attribute::Index index, u8 element = 0) const {
+        const u8 location = static_cast<u8>(static_cast<u32>(index) * 4 + element);
+        const auto it = transform_feedback.find(location);
+        if (it == transform_feedback.end()) {
+            return {};
+        }
+
+        const VaryingTFB& tfb = it->second;
+        return fmt::format("xfb_buffer = {}, xfb_offset = {}, xfb_stride = {}", tfb.buffer,
+                           tfb.offset, tfb.stride);
+    }
+
     void DeclareOutputAttribute(Attribute::Index index) {
-        const u32 location{GetGenericAttributeIndex(index)};
-        code.AddLine("layout (location = {}) out vec4 {};", location, GetOutputAttribute(index));
+        static constexpr std::string_view swizzle = "xyzw";
+        u8 element = 0;
+        while (element < 4) {
+            auto xfb = GetTransformFeedbackDecoration(index, element);
+            if (!xfb.empty()) {
+                xfb = fmt::format(", {}", xfb);
+            }
+            const std::size_t remainder = 4 - element;
+            const std::size_t num_components = GetNumComponents(index, element).value_or(remainder);
+            const char* const type = FLOAT_TYPES.at(num_components - 1);
+
+            const u32 location = GetGenericAttributeIndex(index);
+
+            GenericVaryingDescription description;
+            description.first_element = static_cast<u8>(element);
+            description.is_scalar = num_components == 1;
+            description.name = AppendSuffix(location, OUTPUT_ATTRIBUTE_NAME);
+            if (element != 0 || num_components != 4) {
+                const std::string_view name_swizzle = swizzle.substr(element, num_components);
+                description.name = fmt::format("{}_{}", description.name, name_swizzle);
+            }
+            for (std::size_t i = 0; i < num_components; ++i) {
+                const u8 offset = static_cast<u8>(location * 4 + element + i);
+                varying_description.insert({offset, description});
+            }
+
+            code.AddLine("layout (location = {}, component = {}{}) out {} {};", location, element,
+                         xfb, type, description.name);
+
+            element = static_cast<u8>(static_cast<std::size_t>(element) + num_components);
+        }
     }
 
     void DeclareConstantBuffers() {
@@ -925,7 +1128,8 @@ private:
                 // TODO(Rodrigo): Guard geometry inputs against out of bound reads. Some games
                 // set an 0x80000000 index for those and the shader fails to build. Find out why
                 // this happens and what's its intent.
-                return fmt::format("gs_{}[{} % MAX_VERTEX_INPUT]", name, Visit(buffer).AsUint());
+                return fmt::format("gs_{}[{} % {}]", name, Visit(buffer).AsUint(),
+                                   max_input_vertices.value());
             }
             return std::string(name);
         };
@@ -943,6 +1147,10 @@ private:
             default:
                 UNREACHABLE();
             }
+        case Attribute::Index::FrontColor:
+            return {"gl_Color"s + GetSwizzle(element), Type::Float};
+        case Attribute::Index::FrontSecondaryColor:
+            return {"gl_SecondaryColor"s + GetSwizzle(element), Type::Float};
         case Attribute::Index::PointCoord:
             switch (element) {
             case 0:
@@ -959,7 +1167,7 @@ private:
             // TODO(Subv): Find out what the values are for the first two elements when inside a
             // vertex shader, and what's the value of the fourth element when inside a Tess Eval
             // shader.
-            ASSERT(IsVertexShader(stage));
+            ASSERT(stage == ShaderType::Vertex);
             switch (element) {
             case 2:
                 // Config pack's first value is instance_id.
@@ -980,7 +1188,13 @@ private:
             return {"0", Type::Int};
         default:
             if (IsGenericAttribute(attribute)) {
-                return {GeometryPass(GetInputAttribute(attribute)) + GetSwizzle(element),
+                return {GeometryPass(GetGenericInputAttribute(attribute)) + GetSwizzle(element),
+                        Type::Float};
+            }
+            if (IsLegacyTexCoord(attribute)) {
+                UNIMPLEMENTED_IF(stage == ShaderType::Geometry);
+                return {fmt::format("gl_TexCoord[{}]{}", GetLegacyTexCoordIndex(attribute),
+                                    GetSwizzle(element)),
                         Type::Float};
             }
             break;
@@ -1021,21 +1235,22 @@ private:
     }
 
     std::optional<Expression> GetOutputAttribute(const AbufNode* abuf) {
+        const u32 element = abuf->GetElement();
         switch (const auto attribute = abuf->GetIndex()) {
         case Attribute::Index::Position:
-            return {{"gl_Position"s + GetSwizzle(abuf->GetElement()), Type::Float}};
+            return {{"gl_Position"s + GetSwizzle(element), Type::Float}};
         case Attribute::Index::LayerViewportPointSize:
-            switch (abuf->GetElement()) {
+            switch (element) {
             case 0:
                 UNIMPLEMENTED();
                 return {};
             case 1:
-                if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
+                if (stage == ShaderType::Vertex && !device.HasVertexViewportLayer()) {
                     return {};
                 }
                 return {{"gl_Layer", Type::Int}};
             case 2:
-                if (IsVertexShader(stage) && !device.HasVertexViewportLayer()) {
+                if (stage == ShaderType::Vertex && !device.HasVertexViewportLayer()) {
                     return {};
                 }
                 return {{"gl_ViewportIndex", Type::Int}};
@@ -1043,14 +1258,26 @@ private:
                 return {{"gl_PointSize", Type::Float}};
             }
             return {};
+        case Attribute::Index::FrontColor:
+            return {{"gl_FrontColor"s + GetSwizzle(element), Type::Float}};
+        case Attribute::Index::FrontSecondaryColor:
+            return {{"gl_FrontSecondaryColor"s + GetSwizzle(element), Type::Float}};
+        case Attribute::Index::BackColor:
+            return {{"gl_BackColor"s + GetSwizzle(element), Type::Float}};
+        case Attribute::Index::BackSecondaryColor:
+            return {{"gl_BackSecondaryColor"s + GetSwizzle(element), Type::Float}};
         case Attribute::Index::ClipDistances0123:
-            return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement()), Type::Float}};
+            return {{fmt::format("gl_ClipDistance[{}]", element), Type::Float}};
         case Attribute::Index::ClipDistances4567:
-            return {{fmt::format("gl_ClipDistance[{}]", abuf->GetElement() + 4), Type::Float}};
+            return {{fmt::format("gl_ClipDistance[{}]", element + 4), Type::Float}};
         default:
             if (IsGenericAttribute(attribute)) {
-                return {
-                    {GetOutputAttribute(attribute) + GetSwizzle(abuf->GetElement()), Type::Float}};
+                return {{GetGenericOutputAttribute(attribute, element), Type::Float}};
+            }
+            if (IsLegacyTexCoord(attribute)) {
+                return {{fmt::format("gl_TexCoord[{}]{}", GetLegacyTexCoordIndex(attribute),
+                                     GetSwizzle(element)),
+                         Type::Float}};
             }
             UNIMPLEMENTED_MSG("Unhandled output attribute: {}", static_cast<u32>(attribute));
             return {};
@@ -1822,16 +2049,19 @@ private:
         expr += GetSampler(meta->sampler);
         expr += ", ";
 
-        expr += constructors.at(operation.GetOperandsCount() - 1);
+        expr += constructors.at(operation.GetOperandsCount() + (meta->array ? 1 : 0) - 1);
         expr += '(';
         for (std::size_t i = 0; i < count; ++i) {
-            expr += VisitOperand(operation, i).AsInt();
-            const std::size_t next = i + 1;
-            if (next == count)
-                expr += ')';
-            else if (next < count)
+            if (i > 0) {
                 expr += ", ";
+            }
+            expr += VisitOperand(operation, i).AsInt();
+        }
+        if (meta->array) {
+            expr += ", ";
+            expr += Visit(meta->array).AsInt();
         }
+        expr += ')';
 
         if (meta->lod && !meta->sampler.IsBuffer()) {
             expr += ", ";
@@ -1945,7 +2175,7 @@ private:
             // TODO(Subv): Figure out how dual-source blending is configured in the Switch.
             for (u32 component = 0; component < 4; ++component) {
                 if (header.ps.IsColorComponentOutputEnabled(render_target, component)) {
-                    code.AddLine("FragColor{}[{}] = {};", render_target, component,
+                    code.AddLine("frag_color{}{} = {};", render_target, GetColorSwizzle(component),
                                  SafeGetRegister(current_reg).AsFloat());
                     ++current_reg;
                 }
@@ -2261,27 +2491,34 @@ private:
     static_assert(operation_decompilers.size() == static_cast<std::size_t>(OperationCode::Amount));
 
     std::string GetRegister(u32 index) const {
-        return GetDeclarationWithSuffix(index, "gpr");
+        return AppendSuffix(index, "gpr");
     }
 
     std::string GetCustomVariable(u32 index) const {
-        return GetDeclarationWithSuffix(index, "custom_var");
+        return AppendSuffix(index, "custom_var");
     }
 
     std::string GetPredicate(Tegra::Shader::Pred pred) const {
-        return GetDeclarationWithSuffix(static_cast<u32>(pred), "pred");
+        return AppendSuffix(static_cast<u32>(pred), "pred");
     }
 
-    std::string GetInputAttribute(Attribute::Index attribute) const {
-        return GetDeclarationWithSuffix(GetGenericAttributeIndex(attribute), "input_attr");
+    std::string GetGenericInputAttribute(Attribute::Index attribute) const {
+        return AppendSuffix(GetGenericAttributeIndex(attribute), INPUT_ATTRIBUTE_NAME);
     }
 
-    std::string GetOutputAttribute(Attribute::Index attribute) const {
-        return GetDeclarationWithSuffix(GetGenericAttributeIndex(attribute), "output_attr");
+    std::unordered_map<u8, GenericVaryingDescription> varying_description;
+
+    std::string GetGenericOutputAttribute(Attribute::Index attribute, std::size_t element) const {
+        const u8 offset = static_cast<u8>(GetGenericAttributeIndex(attribute) * 4 + element);
+        const auto& description = varying_description.at(offset);
+        if (description.is_scalar) {
+            return description.name;
+        }
+        return fmt::format("{}[{}]", description.name, element - description.first_element);
     }
 
     std::string GetConstBuffer(u32 index) const {
-        return GetDeclarationWithSuffix(index, "cbuf");
+        return AppendSuffix(index, "cbuf");
     }
 
     std::string GetGlobalMemory(const GlobalMemoryBase& descriptor) const {
@@ -2294,11 +2531,15 @@ private:
     }
 
     std::string GetConstBufferBlock(u32 index) const {
-        return GetDeclarationWithSuffix(index, "cbuf_block");
+        return AppendSuffix(index, "cbuf_block");
     }
 
     std::string GetLocalMemory() const {
-        return "lmem_" + suffix;
+        if (suffix.empty()) {
+            return "lmem";
+        } else {
+            return "lmem_" + std::string{suffix};
+        }
     }
 
     std::string GetInternalFlag(InternalFlag flag) const {
@@ -2307,23 +2548,31 @@ private:
         const auto index = static_cast<u32>(flag);
         ASSERT(index < static_cast<u32>(InternalFlag::Amount));
 
-        return fmt::format("{}_{}", InternalFlagNames[index], suffix);
+        if (suffix.empty()) {
+            return InternalFlagNames[index];
+        } else {
+            return fmt::format("{}_{}", InternalFlagNames[index], suffix);
+        }
     }
 
     std::string GetSampler(const Sampler& sampler) const {
-        return GetDeclarationWithSuffix(static_cast<u32>(sampler.GetIndex()), "sampler");
+        return AppendSuffix(static_cast<u32>(sampler.GetIndex()), "sampler");
     }
 
     std::string GetImage(const Image& image) const {
-        return GetDeclarationWithSuffix(static_cast<u32>(image.GetIndex()), "image");
+        return AppendSuffix(static_cast<u32>(image.GetIndex()), "image");
     }
 
-    std::string GetDeclarationWithSuffix(u32 index, std::string_view name) const {
-        return fmt::format("{}_{}_{}", name, index, suffix);
+    std::string AppendSuffix(u32 index, std::string_view name) const {
+        if (suffix.empty()) {
+            return fmt::format("{}{}", name, index);
+        } else {
+            return fmt::format("{}{}_{}", name, index, suffix);
+        }
     }
 
     u32 GetNumPhysicalInputAttributes() const {
-        return IsVertexShader(stage) ? GetNumPhysicalAttributes() : GetNumPhysicalVaryings();
+        return stage == ShaderType::Vertex ? GetNumPhysicalAttributes() : GetNumPhysicalVaryings();
     }
 
     u32 GetNumPhysicalAttributes() const {
@@ -2334,17 +2583,31 @@ private:
         return std::min<u32>(device.GetMaxVaryings(), Maxwell::NumVaryings);
     }
 
+    bool IsRenderTargetEnabled(u32 render_target) const {
+        for (u32 component = 0; component < 4; ++component) {
+            if (header.ps.IsColorComponentOutputEnabled(render_target, component)) {
+                return true;
+            }
+        }
+        return false;
+    }
+
     const Device& device;
     const ShaderIR& ir;
+    const Registry& registry;
     const ShaderType stage;
-    const std::string suffix;
+    const std::string_view identifier;
+    const std::string_view suffix;
     const Header header;
+    std::unordered_map<u8, VaryingTFB> transform_feedback;
 
     ShaderWriter code;
+
+    std::optional<u32> max_input_vertices;
 };
 
-std::string GetFlowVariable(u32 i) {
-    return fmt::format("flow_var_{}", i);
+std::string GetFlowVariable(u32 index) {
+    return fmt::format("flow_var{}", index);
 }
 
 class ExprDecompiler {
@@ -2531,7 +2794,7 @@ void GLSLDecompiler::DecompileAST() {
 
 } // Anonymous namespace
 
-ShaderEntries GetEntries(const VideoCommon::Shader::ShaderIR& ir) {
+ShaderEntries MakeEntries(const VideoCommon::Shader::ShaderIR& ir) {
     ShaderEntries entries;
     for (const auto& cbuf : ir.GetConstantBuffers()) {
         entries.const_buffers.emplace_back(cbuf.second.GetMaxOffset(), cbuf.second.IsIndirect(),
@@ -2547,33 +2810,20 @@ ShaderEntries GetEntries(const VideoCommon::Shader::ShaderIR& ir) {
     for (const auto& image : ir.GetImages()) {
         entries.images.emplace_back(image);
     }
-    entries.clip_distances = ir.GetClipDistances();
+    const auto clip_distances = ir.GetClipDistances();
+    for (std::size_t i = 0; i < std::size(clip_distances); ++i) {
+        entries.clip_distances = (clip_distances[i] ? 1U : 0U) << i;
+    }
     entries.shader_length = ir.GetLength();
     return entries;
 }
 
-std::string GetCommonDeclarations() {
-    return R"(#define ftoi floatBitsToInt
-#define ftou floatBitsToUint
-#define itof intBitsToFloat
-#define utof uintBitsToFloat
-
-bvec2 HalfFloatNanComparison(bvec2 comparison, vec2 pair1, vec2 pair2) {
-    bvec2 is_nan1 = isnan(pair1);
-    bvec2 is_nan2 = isnan(pair2);
-    return bvec2(comparison.x || is_nan1.x || is_nan2.x, comparison.y || is_nan1.y || is_nan2.y);
-}
-
-const float fswzadd_modifiers_a[] = float[4](-1.0f,  1.0f, -1.0f,  0.0f );
-const float fswzadd_modifiers_b[] = float[4](-1.0f, -1.0f,  1.0f, -1.0f );
-)";
-}
-
-std::string Decompile(const Device& device, const ShaderIR& ir, ShaderType stage,
-                      const std::string& suffix) {
-    GLSLDecompiler decompiler(device, ir, stage, suffix);
+std::string DecompileShader(const Device& device, const ShaderIR& ir, const Registry& registry,
+                            ShaderType stage, std::string_view identifier,
+                            std::string_view suffix) {
+    GLSLDecompiler decompiler(device, ir, registry, stage, identifier, suffix);
     decompiler.Decompile();
     return decompiler.GetResult();
 }
 
-} // namespace OpenGL::GLShader
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.h b/src/video_core/renderer_opengl/gl_shader_decompiler.h
index 7876f48d6..e7dbd810c 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.h
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.h
@@ -6,22 +6,18 @@
 
 #include <array>
 #include <string>
+#include <string_view>
 #include <utility>
 #include <vector>
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/shader_type.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
-namespace VideoCommon::Shader {
-class ShaderIR;
-}
-
 namespace OpenGL {
-class Device;
-}
 
-namespace OpenGL::GLShader {
+class Device;
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 using SamplerEntry = VideoCommon::Shader::Sampler;
@@ -74,15 +70,15 @@ struct ShaderEntries {
     std::vector<GlobalMemoryEntry> global_memory_entries;
     std::vector<SamplerEntry> samplers;
     std::vector<ImageEntry> images;
-    std::array<bool, Maxwell::NumClipDistances> clip_distances{};
+    u32 clip_distances{};
     std::size_t shader_length{};
 };
 
-ShaderEntries GetEntries(const VideoCommon::Shader::ShaderIR& ir);
-
-std::string GetCommonDeclarations();
+ShaderEntries MakeEntries(const VideoCommon::Shader::ShaderIR& ir);
 
-std::string Decompile(const Device& device, const VideoCommon::Shader::ShaderIR& ir,
-                      Tegra::Engines::ShaderType stage, const std::string& suffix);
+std::string DecompileShader(const Device& device, const VideoCommon::Shader::ShaderIR& ir,
+                            const VideoCommon::Shader::Registry& registry,
+                            Tegra::Engines::ShaderType stage, std::string_view identifier,
+                            std::string_view suffix = {});
 
-} // namespace OpenGL::GLShader
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
index 1fc204f6f..9e95a122b 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -31,32 +31,24 @@ namespace {
 
 using ShaderCacheVersionHash = std::array<u8, 64>;
 
-enum class TransferableEntryKind : u32 {
-    Raw,
-    Usage,
-};
-
 struct ConstBufferKey {
-    u32 cbuf{};
-    u32 offset{};
-    u32 value{};
+    u32 cbuf = 0;
+    u32 offset = 0;
+    u32 value = 0;
 };
 
 struct BoundSamplerKey {
-    u32 offset{};
-    Tegra::Engines::SamplerDescriptor sampler{};
+    u32 offset = 0;
+    Tegra::Engines::SamplerDescriptor sampler;
 };
 
 struct BindlessSamplerKey {
-    u32 cbuf{};
-    u32 offset{};
-    Tegra::Engines::SamplerDescriptor sampler{};
+    u32 cbuf = 0;
+    u32 offset = 0;
+    Tegra::Engines::SamplerDescriptor sampler;
 };
 
-constexpr u32 NativeVersion = 12;
-
-// Making sure sizes doesn't change by accident
-static_assert(sizeof(ProgramVariant) == 20);
+constexpr u32 NativeVersion = 20;
 
 ShaderCacheVersionHash GetShaderCacheVersionHash() {
     ShaderCacheVersionHash hash{};
@@ -67,61 +59,124 @@ ShaderCacheVersionHash GetShaderCacheVersionHash() {
 
 } // Anonymous namespace
 
-ShaderDiskCacheRaw::ShaderDiskCacheRaw(u64 unique_identifier, ShaderType type, ProgramCode code,
-                                       ProgramCode code_b)
-    : unique_identifier{unique_identifier}, type{type}, code{std::move(code)}, code_b{std::move(
-                                                                                   code_b)} {}
+ShaderDiskCacheEntry::ShaderDiskCacheEntry() = default;
 
-ShaderDiskCacheRaw::ShaderDiskCacheRaw() = default;
+ShaderDiskCacheEntry::~ShaderDiskCacheEntry() = default;
 
-ShaderDiskCacheRaw::~ShaderDiskCacheRaw() = default;
-
-bool ShaderDiskCacheRaw::Load(FileUtil::IOFile& file) {
-    if (file.ReadBytes(&unique_identifier, sizeof(u64)) != sizeof(u64) ||
-        file.ReadBytes(&type, sizeof(u32)) != sizeof(u32)) {
+bool ShaderDiskCacheEntry::Load(FileUtil::IOFile& file) {
+    if (file.ReadBytes(&type, sizeof(u32)) != sizeof(u32)) {
         return false;
     }
-    u32 code_size{};
-    u32 code_size_b{};
+    u32 code_size;
+    u32 code_size_b;
     if (file.ReadBytes(&code_size, sizeof(u32)) != sizeof(u32) ||
         file.ReadBytes(&code_size_b, sizeof(u32)) != sizeof(u32)) {
         return false;
     }
-
     code.resize(code_size);
     code_b.resize(code_size_b);
 
-    if (file.ReadArray(code.data(), code_size) != code_size)
+    if (file.ReadArray(code.data(), code_size) != code_size) {
         return false;
-
+    }
     if (HasProgramA() && file.ReadArray(code_b.data(), code_size_b) != code_size_b) {
         return false;
     }
+
+    u8 is_texture_handler_size_known;
+    u32 texture_handler_size_value;
+    u32 num_keys;
+    u32 num_bound_samplers;
+    u32 num_bindless_samplers;
+    if (file.ReadArray(&unique_identifier, 1) != 1 || file.ReadArray(&bound_buffer, 1) != 1 ||
+        file.ReadArray(&is_texture_handler_size_known, 1) != 1 ||
+        file.ReadArray(&texture_handler_size_value, 1) != 1 ||
+        file.ReadArray(&graphics_info, 1) != 1 || file.ReadArray(&compute_info, 1) != 1 ||
+        file.ReadArray(&num_keys, 1) != 1 || file.ReadArray(&num_bound_samplers, 1) != 1 ||
+        file.ReadArray(&num_bindless_samplers, 1) != 1) {
+        return false;
+    }
+    if (is_texture_handler_size_known) {
+        texture_handler_size = texture_handler_size_value;
+    }
+
+    std::vector<ConstBufferKey> flat_keys(num_keys);
+    std::vector<BoundSamplerKey> flat_bound_samplers(num_bound_samplers);
+    std::vector<BindlessSamplerKey> flat_bindless_samplers(num_bindless_samplers);
+    if (file.ReadArray(flat_keys.data(), flat_keys.size()) != flat_keys.size() ||
+        file.ReadArray(flat_bound_samplers.data(), flat_bound_samplers.size()) !=
+            flat_bound_samplers.size() ||
+        file.ReadArray(flat_bindless_samplers.data(), flat_bindless_samplers.size()) !=
+            flat_bindless_samplers.size()) {
+        return false;
+    }
+    for (const auto& key : flat_keys) {
+        keys.insert({{key.cbuf, key.offset}, key.value});
+    }
+    for (const auto& key : flat_bound_samplers) {
+        bound_samplers.emplace(key.offset, key.sampler);
+    }
+    for (const auto& key : flat_bindless_samplers) {
+        bindless_samplers.insert({{key.cbuf, key.offset}, key.sampler});
+    }
+
     return true;
 }
 
-bool ShaderDiskCacheRaw::Save(FileUtil::IOFile& file) const {
-    if (file.WriteObject(unique_identifier) != 1 || file.WriteObject(static_cast<u32>(type)) != 1 ||
+bool ShaderDiskCacheEntry::Save(FileUtil::IOFile& file) const {
+    if (file.WriteObject(static_cast<u32>(type)) != 1 ||
         file.WriteObject(static_cast<u32>(code.size())) != 1 ||
         file.WriteObject(static_cast<u32>(code_b.size())) != 1) {
         return false;
     }
-
-    if (file.WriteArray(code.data(), code.size()) != code.size())
+    if (file.WriteArray(code.data(), code.size()) != code.size()) {
         return false;
-
+    }
     if (HasProgramA() && file.WriteArray(code_b.data(), code_b.size()) != code_b.size()) {
         return false;
     }
-    return true;
+
+    if (file.WriteObject(unique_identifier) != 1 || file.WriteObject(bound_buffer) != 1 ||
+        file.WriteObject(static_cast<u8>(texture_handler_size.has_value())) != 1 ||
+        file.WriteObject(texture_handler_size.value_or(0)) != 1 ||
+        file.WriteObject(graphics_info) != 1 || file.WriteObject(compute_info) != 1 ||
+        file.WriteObject(static_cast<u32>(keys.size())) != 1 ||
+        file.WriteObject(static_cast<u32>(bound_samplers.size())) != 1 ||
+        file.WriteObject(static_cast<u32>(bindless_samplers.size())) != 1) {
+        return false;
+    }
+
+    std::vector<ConstBufferKey> flat_keys;
+    flat_keys.reserve(keys.size());
+    for (const auto& [address, value] : keys) {
+        flat_keys.push_back(ConstBufferKey{address.first, address.second, value});
+    }
+
+    std::vector<BoundSamplerKey> flat_bound_samplers;
+    flat_bound_samplers.reserve(bound_samplers.size());
+    for (const auto& [address, sampler] : bound_samplers) {
+        flat_bound_samplers.push_back(BoundSamplerKey{address, sampler});
+    }
+
+    std::vector<BindlessSamplerKey> flat_bindless_samplers;
+    flat_bindless_samplers.reserve(bindless_samplers.size());
+    for (const auto& [address, sampler] : bindless_samplers) {
+        flat_bindless_samplers.push_back(
+            BindlessSamplerKey{address.first, address.second, sampler});
+    }
+
+    return file.WriteArray(flat_keys.data(), flat_keys.size()) == flat_keys.size() &&
+           file.WriteArray(flat_bound_samplers.data(), flat_bound_samplers.size()) ==
+               flat_bound_samplers.size() &&
+           file.WriteArray(flat_bindless_samplers.data(), flat_bindless_samplers.size()) ==
+               flat_bindless_samplers.size();
 }
 
 ShaderDiskCacheOpenGL::ShaderDiskCacheOpenGL(Core::System& system) : system{system} {}
 
 ShaderDiskCacheOpenGL::~ShaderDiskCacheOpenGL() = default;
 
-std::optional<std::pair<std::vector<ShaderDiskCacheRaw>, std::vector<ShaderDiskCacheUsage>>>
-ShaderDiskCacheOpenGL::LoadTransferable() {
+std::optional<std::vector<ShaderDiskCacheEntry>> 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) {
@@ -130,17 +185,14 @@ ShaderDiskCacheOpenGL::LoadTransferable() {
 
     FileUtil::IOFile file(GetTransferablePath(), "rb");
     if (!file.IsOpen()) {
-        LOG_INFO(Render_OpenGL, "No transferable shader cache found for game with title id={}",
-                 GetTitleID());
+        LOG_INFO(Render_OpenGL, "No transferable shader cache found");
         is_usable = true;
         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());
+        LOG_ERROR(Render_OpenGL, "Failed to get transferable cache version, skipping it");
         return {};
     }
 
@@ -158,105 +210,42 @@ ShaderDiskCacheOpenGL::LoadTransferable() {
     }
 
     // Version is valid, load the shaders
-    constexpr const char error_loading[] = "Failed to load transferable raw entry, skipping";
-    std::vector<ShaderDiskCacheRaw> raws;
-    std::vector<ShaderDiskCacheUsage> usages;
+    std::vector<ShaderDiskCacheEntry> entries;
     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, error_loading);
-                return {};
-            }
-            transferable.insert({entry.GetUniqueIdentifier(), {}});
-            raws.push_back(std::move(entry));
-            break;
-        }
-        case TransferableEntryKind::Usage: {
-            ShaderDiskCacheUsage usage;
-
-            u32 num_keys{};
-            u32 num_bound_samplers{};
-            u32 num_bindless_samplers{};
-            if (file.ReadArray(&usage.unique_identifier, 1) != 1 ||
-                file.ReadArray(&usage.variant, 1) != 1 ||
-                file.ReadArray(&usage.bound_buffer, 1) != 1 || file.ReadArray(&num_keys, 1) != 1 ||
-                file.ReadArray(&num_bound_samplers, 1) != 1 ||
-                file.ReadArray(&num_bindless_samplers, 1) != 1) {
-                LOG_ERROR(Render_OpenGL, error_loading);
-                return {};
-            }
-
-            std::vector<ConstBufferKey> keys(num_keys);
-            std::vector<BoundSamplerKey> bound_samplers(num_bound_samplers);
-            std::vector<BindlessSamplerKey> bindless_samplers(num_bindless_samplers);
-            if (file.ReadArray(keys.data(), keys.size()) != keys.size() ||
-                file.ReadArray(bound_samplers.data(), bound_samplers.size()) !=
-                    bound_samplers.size() ||
-                file.ReadArray(bindless_samplers.data(), bindless_samplers.size()) !=
-                    bindless_samplers.size()) {
-                LOG_ERROR(Render_OpenGL, error_loading);
-                return {};
-            }
-            for (const auto& key : keys) {
-                usage.keys.insert({{key.cbuf, key.offset}, key.value});
-            }
-            for (const auto& key : bound_samplers) {
-                usage.bound_samplers.emplace(key.offset, key.sampler);
-            }
-            for (const auto& key : bindless_samplers) {
-                usage.bindless_samplers.insert({{key.cbuf, key.offset}, key.sampler});
-            }
-
-            usages.push_back(std::move(usage));
-            break;
-        }
-        default:
-            LOG_ERROR(Render_OpenGL, "Unknown transferable shader cache entry kind={}, skipping",
-                      static_cast<u32>(kind));
+        ShaderDiskCacheEntry& entry = entries.emplace_back();
+        if (!entry.Load(file)) {
+            LOG_ERROR(Render_OpenGL, "Failed to load transferable raw entry, skipping");
             return {};
         }
     }
 
     is_usable = true;
-    return {{std::move(raws), std::move(usages)}};
+    return {std::move(entries)};
 }
 
-std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>
-ShaderDiskCacheOpenGL::LoadPrecompiled() {
+std::vector<ShaderDiskCachePrecompiled> ShaderDiskCacheOpenGL::LoadPrecompiled() {
     if (!is_usable) {
         return {};
     }
 
-    std::string path = GetPrecompiledPath();
-    FileUtil::IOFile file(path, "rb");
+    FileUtil::IOFile file(GetPrecompiledPath(), "rb");
     if (!file.IsOpen()) {
-        LOG_INFO(Render_OpenGL, "No precompiled shader cache found for game with title id={}",
-                 GetTitleID());
+        LOG_INFO(Render_OpenGL, "No precompiled shader cache found");
         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 {};
+    if (const auto result = LoadPrecompiledFile(file)) {
+        return *result;
     }
-    return *result;
+
+    LOG_INFO(Render_OpenGL, "Failed to load precompiled cache");
+    file.Close();
+    InvalidatePrecompiled();
+    return {};
 }
 
-std::optional<std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>
-ShaderDiskCacheOpenGL::LoadPrecompiledFile(FileUtil::IOFile& file) {
+std::optional<std::vector<ShaderDiskCachePrecompiled>> ShaderDiskCacheOpenGL::LoadPrecompiledFile(
+    FileUtil::IOFile& file) {
     // Read compressed file from disk and decompress to virtual precompiled cache file
     std::vector<u8> compressed(file.GetSize());
     file.ReadBytes(compressed.data(), compressed.size());
@@ -275,58 +264,22 @@ ShaderDiskCacheOpenGL::LoadPrecompiledFile(FileUtil::IOFile& file) {
         return {};
     }
 
-    ShaderDumpsMap dumps;
+    std::vector<ShaderDiskCachePrecompiled> entries;
     while (precompiled_cache_virtual_file_offset < precompiled_cache_virtual_file.GetSize()) {
-        u32 num_keys{};
-        u32 num_bound_samplers{};
-        u32 num_bindless_samplers{};
-        ShaderDiskCacheUsage usage;
-        if (!LoadObjectFromPrecompiled(usage.unique_identifier) ||
-            !LoadObjectFromPrecompiled(usage.variant) ||
-            !LoadObjectFromPrecompiled(usage.bound_buffer) ||
-            !LoadObjectFromPrecompiled(num_keys) ||
-            !LoadObjectFromPrecompiled(num_bound_samplers) ||
-            !LoadObjectFromPrecompiled(num_bindless_samplers)) {
-            return {};
-        }
-        std::vector<ConstBufferKey> keys(num_keys);
-        std::vector<BoundSamplerKey> bound_samplers(num_bound_samplers);
-        std::vector<BindlessSamplerKey> bindless_samplers(num_bindless_samplers);
-        if (!LoadArrayFromPrecompiled(keys.data(), keys.size()) ||
-            !LoadArrayFromPrecompiled(bound_samplers.data(), bound_samplers.size()) !=
-                bound_samplers.size() ||
-            !LoadArrayFromPrecompiled(bindless_samplers.data(), bindless_samplers.size()) !=
-                bindless_samplers.size()) {
-            return {};
-        }
-        for (const auto& key : keys) {
-            usage.keys.insert({{key.cbuf, key.offset}, key.value});
-        }
-        for (const auto& key : bound_samplers) {
-            usage.bound_samplers.emplace(key.offset, key.sampler);
-        }
-        for (const auto& key : bindless_samplers) {
-            usage.bindless_samplers.insert({{key.cbuf, key.offset}, key.sampler});
-        }
-
-        ShaderDiskCacheDump dump;
-        if (!LoadObjectFromPrecompiled(dump.binary_format)) {
-            return {};
-        }
-
-        u32 binary_length{};
-        if (!LoadObjectFromPrecompiled(binary_length)) {
+        u32 binary_size;
+        auto& entry = entries.emplace_back();
+        if (!LoadObjectFromPrecompiled(entry.unique_identifier) ||
+            !LoadObjectFromPrecompiled(entry.binary_format) ||
+            !LoadObjectFromPrecompiled(binary_size)) {
             return {};
         }
 
-        dump.binary.resize(binary_length);
-        if (!LoadArrayFromPrecompiled(dump.binary.data(), dump.binary.size())) {
+        entry.binary.resize(binary_size);
+        if (!LoadArrayFromPrecompiled(entry.binary.data(), entry.binary.size())) {
             return {};
         }
-
-        dumps.emplace(std::move(usage), dump);
     }
-    return dumps;
+    return entries;
 }
 
 void ShaderDiskCacheOpenGL::InvalidateTransferable() {
@@ -346,13 +299,13 @@ void ShaderDiskCacheOpenGL::InvalidatePrecompiled() {
     }
 }
 
-void ShaderDiskCacheOpenGL::SaveRaw(const ShaderDiskCacheRaw& entry) {
+void ShaderDiskCacheOpenGL::SaveEntry(const ShaderDiskCacheEntry& entry) {
     if (!is_usable) {
         return;
     }
 
-    const u64 id = entry.GetUniqueIdentifier();
-    if (transferable.find(id) != transferable.end()) {
+    const u64 id = entry.unique_identifier;
+    if (stored_transferable.find(id) != stored_transferable.end()) {
         // The shader already exists
         return;
     }
@@ -361,71 +314,17 @@ void ShaderDiskCacheOpenGL::SaveRaw(const ShaderDiskCacheRaw& entry) {
     if (!file.IsOpen()) {
         return;
     }
-    if (file.WriteObject(TransferableEntryKind::Raw) != 1 || !entry.Save(file)) {
+    if (!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 (!is_usable) {
-        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};
-    if (usages.find(usage) != usages.end()) {
-        // Skip this variant since the shader is already stored.
-        return;
-    }
-    usages.insert(usage);
-
-    FileUtil::IOFile file = AppendTransferableFile();
-    if (!file.IsOpen())
-        return;
-    const auto Close = [&] {
-        LOG_ERROR(Render_OpenGL, "Failed to save usage transferable cache entry, removing");
-        file.Close();
-        InvalidateTransferable();
-    };
-
-    if (file.WriteObject(TransferableEntryKind::Usage) != 1 ||
-        file.WriteObject(usage.unique_identifier) != 1 || file.WriteObject(usage.variant) != 1 ||
-        file.WriteObject(usage.bound_buffer) != 1 ||
-        file.WriteObject(static_cast<u32>(usage.keys.size())) != 1 ||
-        file.WriteObject(static_cast<u32>(usage.bound_samplers.size())) != 1 ||
-        file.WriteObject(static_cast<u32>(usage.bindless_samplers.size())) != 1) {
-        Close();
-        return;
-    }
-    for (const auto& [pair, value] : usage.keys) {
-        const auto [cbuf, offset] = pair;
-        if (file.WriteObject(ConstBufferKey{cbuf, offset, value}) != 1) {
-            Close();
-            return;
-        }
-    }
-    for (const auto& [offset, sampler] : usage.bound_samplers) {
-        if (file.WriteObject(BoundSamplerKey{offset, sampler}) != 1) {
-            Close();
-            return;
-        }
-    }
-    for (const auto& [pair, sampler] : usage.bindless_samplers) {
-        const auto [cbuf, offset] = pair;
-        if (file.WriteObject(BindlessSamplerKey{cbuf, offset, sampler}) != 1) {
-            Close();
-            return;
-        }
-    }
+    stored_transferable.insert(id);
 }
 
-void ShaderDiskCacheOpenGL::SaveDump(const ShaderDiskCacheUsage& usage, GLuint program) {
+void ShaderDiskCacheOpenGL::SavePrecompiled(u64 unique_identifier, GLuint program) {
     if (!is_usable) {
         return;
     }
@@ -437,51 +336,19 @@ void ShaderDiskCacheOpenGL::SaveDump(const ShaderDiskCacheUsage& usage, GLuint p
         SavePrecompiledHeaderToVirtualPrecompiledCache();
     }
 
-    GLint binary_length{};
+    GLint binary_length;
     glGetProgramiv(program, GL_PROGRAM_BINARY_LENGTH, &binary_length);
 
-    GLenum binary_format{};
+    GLenum binary_format;
     std::vector<u8> binary(binary_length);
     glGetProgramBinary(program, binary_length, nullptr, &binary_format, binary.data());
 
-    const auto Close = [&] {
+    if (!SaveObjectToPrecompiled(unique_identifier) || !SaveObjectToPrecompiled(binary_format) ||
+        !SaveObjectToPrecompiled(static_cast<u32>(binary.size())) ||
+        !SaveArrayToPrecompiled(binary.data(), binary.size())) {
         LOG_ERROR(Render_OpenGL, "Failed to save binary program file in shader={:016X}, removing",
-                  usage.unique_identifier);
+                  unique_identifier);
         InvalidatePrecompiled();
-    };
-
-    if (!SaveObjectToPrecompiled(usage.unique_identifier) ||
-        !SaveObjectToPrecompiled(usage.variant) || !SaveObjectToPrecompiled(usage.bound_buffer) ||
-        !SaveObjectToPrecompiled(static_cast<u32>(usage.keys.size())) ||
-        !SaveObjectToPrecompiled(static_cast<u32>(usage.bound_samplers.size())) ||
-        !SaveObjectToPrecompiled(static_cast<u32>(usage.bindless_samplers.size()))) {
-        Close();
-        return;
-    }
-    for (const auto& [pair, value] : usage.keys) {
-        const auto [cbuf, offset] = pair;
-        if (SaveObjectToPrecompiled(ConstBufferKey{cbuf, offset, value}) != 1) {
-            Close();
-            return;
-        }
-    }
-    for (const auto& [offset, sampler] : usage.bound_samplers) {
-        if (SaveObjectToPrecompiled(BoundSamplerKey{offset, sampler}) != 1) {
-            Close();
-            return;
-        }
-    }
-    for (const auto& [pair, sampler] : usage.bindless_samplers) {
-        const auto [cbuf, offset] = pair;
-        if (SaveObjectToPrecompiled(BindlessSamplerKey{cbuf, offset, sampler}) != 1) {
-            Close();
-            return;
-        }
-    }
-    if (!SaveObjectToPrecompiled(static_cast<u32>(binary_format)) ||
-        !SaveObjectToPrecompiled(static_cast<u32>(binary_length)) ||
-        !SaveArrayToPrecompiled(binary.data(), binary.size())) {
-        Close();
     }
 }
 
@@ -534,7 +401,6 @@ void ShaderDiskCacheOpenGL::SaveVirtualPrecompiledFile() {
     if (file.WriteBytes(compressed.data(), compressed.size()) != compressed.size()) {
         LOG_ERROR(Render_OpenGL, "Failed to write precompiled cache version in path={}",
                   precompiled_path);
-        return;
     }
 }
 
diff --git a/src/video_core/renderer_opengl/gl_shader_disk_cache.h b/src/video_core/renderer_opengl/gl_shader_disk_cache.h
index ef2371f6d..d5be52e40 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.h
@@ -19,8 +19,7 @@
 #include "common/common_types.h"
 #include "core/file_sys/vfs_vector.h"
 #include "video_core/engines/shader_type.h"
-#include "video_core/renderer_opengl/gl_shader_gen.h"
-#include "video_core/shader/const_buffer_locker.h"
+#include "video_core/shader/registry.h"
 
 namespace Core {
 class System;
@@ -32,139 +31,39 @@ class IOFile;
 
 namespace OpenGL {
 
-struct ShaderDiskCacheUsage;
-struct ShaderDiskCacheDump;
-
 using ProgramCode = std::vector<u64>;
-using ShaderDumpsMap = std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>;
-
-/// Describes the different variants a program can be compiled with.
-struct ProgramVariant final {
-    ProgramVariant() = default;
-
-    /// Graphics constructor.
-    explicit constexpr ProgramVariant(GLenum primitive_mode) noexcept
-        : primitive_mode{primitive_mode} {}
-
-    /// Compute constructor.
-    explicit constexpr ProgramVariant(u32 block_x, u32 block_y, u32 block_z, u32 shared_memory_size,
-                                      u32 local_memory_size) noexcept
-        : block_x{block_x}, block_y{static_cast<u16>(block_y)}, block_z{static_cast<u16>(block_z)},
-          shared_memory_size{shared_memory_size}, local_memory_size{local_memory_size} {}
-
-    // Graphics specific parameters.
-    GLenum primitive_mode{};
-
-    // Compute specific parameters.
-    u32 block_x{};
-    u16 block_y{};
-    u16 block_z{};
-    u32 shared_memory_size{};
-    u32 local_memory_size{};
-
-    bool operator==(const ProgramVariant& rhs) const noexcept {
-        return std::tie(primitive_mode, block_x, block_y, block_z, shared_memory_size,
-                        local_memory_size) == std::tie(rhs.primitive_mode, rhs.block_x, rhs.block_y,
-                                                       rhs.block_z, rhs.shared_memory_size,
-                                                       rhs.local_memory_size);
-    }
-
-    bool operator!=(const ProgramVariant& rhs) const noexcept {
-        return !operator==(rhs);
-    }
-};
-static_assert(std::is_trivially_copyable_v<ProgramVariant>);
-
-/// Describes how a shader is used.
-struct ShaderDiskCacheUsage {
-    u64 unique_identifier{};
-    ProgramVariant variant;
-    u32 bound_buffer{};
-    VideoCommon::Shader::KeyMap keys;
-    VideoCommon::Shader::BoundSamplerMap bound_samplers;
-    VideoCommon::Shader::BindlessSamplerMap bindless_samplers;
-
-    bool operator==(const ShaderDiskCacheUsage& rhs) const {
-        return std::tie(unique_identifier, variant, keys, bound_samplers, bindless_samplers) ==
-               std::tie(rhs.unique_identifier, rhs.variant, rhs.keys, rhs.bound_samplers,
-                        rhs.bindless_samplers);
-    }
-
-    bool operator!=(const ShaderDiskCacheUsage& rhs) const {
-        return !operator==(rhs);
-    }
-};
-
-} // namespace OpenGL
-
-namespace std {
-
-template <>
-struct hash<OpenGL::ProgramVariant> {
-    std::size_t operator()(const OpenGL::ProgramVariant& variant) const noexcept {
-        return (static_cast<std::size_t>(variant.primitive_mode) << 6) ^
-               static_cast<std::size_t>(variant.block_x) ^
-               (static_cast<std::size_t>(variant.block_y) << 32) ^
-               (static_cast<std::size_t>(variant.block_z) << 48) ^
-               (static_cast<std::size_t>(variant.shared_memory_size) << 16) ^
-               (static_cast<std::size_t>(variant.local_memory_size) << 36);
-    }
-};
-
-template <>
-struct hash<OpenGL::ShaderDiskCacheUsage> {
-    std::size_t operator()(const OpenGL::ShaderDiskCacheUsage& usage) const noexcept {
-        return static_cast<std::size_t>(usage.unique_identifier) ^
-               std::hash<OpenGL::ProgramVariant>{}(usage.variant);
-    }
-};
-
-} // namespace std
-
-namespace OpenGL {
 
-/// Describes a shader how it's used by the guest GPU
-class ShaderDiskCacheRaw {
-public:
-    explicit ShaderDiskCacheRaw(u64 unique_identifier, Tegra::Engines::ShaderType type,
-                                ProgramCode code, ProgramCode code_b = {});
-    ShaderDiskCacheRaw();
-    ~ShaderDiskCacheRaw();
+/// Describes a shader and how it's used by the guest GPU
+struct ShaderDiskCacheEntry {
+    ShaderDiskCacheEntry();
+    ~ShaderDiskCacheEntry();
 
     bool Load(FileUtil::IOFile& file);
 
     bool Save(FileUtil::IOFile& file) const;
 
-    u64 GetUniqueIdentifier() const {
-        return unique_identifier;
-    }
-
     bool HasProgramA() const {
         return !code.empty() && !code_b.empty();
     }
 
-    Tegra::Engines::ShaderType GetType() const {
-        return type;
-    }
-
-    const ProgramCode& GetCode() const {
-        return code;
-    }
-
-    const ProgramCode& GetCodeB() const {
-        return code_b;
-    }
-
-private:
-    u64 unique_identifier{};
     Tegra::Engines::ShaderType type{};
     ProgramCode code;
     ProgramCode code_b;
+
+    u64 unique_identifier = 0;
+    std::optional<u32> texture_handler_size;
+    u32 bound_buffer = 0;
+    VideoCommon::Shader::GraphicsInfo graphics_info;
+    VideoCommon::Shader::ComputeInfo compute_info;
+    VideoCommon::Shader::KeyMap keys;
+    VideoCommon::Shader::BoundSamplerMap bound_samplers;
+    VideoCommon::Shader::BindlessSamplerMap bindless_samplers;
 };
 
 /// Contains an OpenGL dumped binary program
-struct ShaderDiskCacheDump {
-    GLenum binary_format{};
+struct ShaderDiskCachePrecompiled {
+    u64 unique_identifier = 0;
+    GLenum binary_format = 0;
     std::vector<u8> binary;
 };
 
@@ -174,11 +73,10 @@ public:
     ~ShaderDiskCacheOpenGL();
 
     /// 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();
+    std::optional<std::vector<ShaderDiskCacheEntry>> LoadTransferable();
 
     /// Loads current game's precompiled cache. Invalidates on failure.
-    std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump> LoadPrecompiled();
+    std::vector<ShaderDiskCachePrecompiled> LoadPrecompiled();
 
     /// Removes the transferable (and precompiled) cache file.
     void InvalidateTransferable();
@@ -187,21 +85,18 @@ public:
     void InvalidatePrecompiled();
 
     /// 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);
+    void SaveEntry(const ShaderDiskCacheEntry& entry);
 
     /// Saves a dump entry to the precompiled file. Does not check for collisions.
-    void SaveDump(const ShaderDiskCacheUsage& usage, GLuint program);
+    void SavePrecompiled(u64 unique_identifier, GLuint program);
 
     /// Serializes virtual precompiled shader cache file to real file
     void SaveVirtualPrecompiledFile();
 
 private:
     /// Loads the transferable cache. Returns empty on failure.
-    std::optional<std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>
-    LoadPrecompiledFile(FileUtil::IOFile& file);
+    std::optional<std::vector<ShaderDiskCachePrecompiled>> LoadPrecompiledFile(
+        FileUtil::IOFile& file);
 
     /// Opens current game's transferable file and write it's header if it doesn't exist
     FileUtil::IOFile AppendTransferableFile() const;
@@ -270,7 +165,7 @@ private:
     std::size_t precompiled_cache_virtual_file_offset = 0;
 
     // Stored transferable shaders
-    std::unordered_map<u64, std::unordered_set<ShaderDiskCacheUsage>> transferable;
+    std::unordered_set<u64> stored_transferable;
 
     // The cache has been loaded at boot
     bool is_usable{};
diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp
deleted file mode 100644
index 34946fb47..000000000
--- a/src/video_core/renderer_opengl/gl_shader_gen.cpp
+++ /dev/null
@@ -1,109 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <string>
-
-#include <fmt/format.h>
-
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/engines/shader_type.h"
-#include "video_core/renderer_opengl/gl_device.h"
-#include "video_core/renderer_opengl/gl_shader_decompiler.h"
-#include "video_core/renderer_opengl/gl_shader_gen.h"
-#include "video_core/shader/shader_ir.h"
-
-namespace OpenGL::GLShader {
-
-using Tegra::Engines::Maxwell3D;
-using Tegra::Engines::ShaderType;
-using VideoCommon::Shader::CompileDepth;
-using VideoCommon::Shader::CompilerSettings;
-using VideoCommon::Shader::ProgramCode;
-using VideoCommon::Shader::ShaderIR;
-
-std::string GenerateVertexShader(const Device& device, const ShaderIR& ir, const ShaderIR* ir_b) {
-    std::string out = GetCommonDeclarations();
-    out += fmt::format(R"(
-layout (std140, binding = {}) uniform vs_config {{
-    float y_direction;
-}};
-
-)",
-                       EmulationUniformBlockBinding);
-    out += Decompile(device, ir, ShaderType::Vertex, "vertex");
-    if (ir_b) {
-        out += Decompile(device, *ir_b, ShaderType::Vertex, "vertex_b");
-    }
-
-    out += R"(
-void main() {
-    gl_Position = vec4(0.0f, 0.0f, 0.0f, 1.0f);
-    execute_vertex();
-)";
-    if (ir_b) {
-        out += "    execute_vertex_b();";
-    }
-    out += "}\n";
-    return out;
-}
-
-std::string GenerateGeometryShader(const Device& device, const ShaderIR& ir) {
-    std::string out = GetCommonDeclarations();
-    out += fmt::format(R"(
-layout (std140, binding = {}) uniform gs_config {{
-    float y_direction;
-}};
-
-)",
-                       EmulationUniformBlockBinding);
-    out += Decompile(device, ir, ShaderType::Geometry, "geometry");
-
-    out += R"(
-void main() {
-    execute_geometry();
-}
-)";
-    return out;
-}
-
-std::string GenerateFragmentShader(const Device& device, const ShaderIR& ir) {
-    std::string out = GetCommonDeclarations();
-    out += fmt::format(R"(
-layout (location = 0) out vec4 FragColor0;
-layout (location = 1) out vec4 FragColor1;
-layout (location = 2) out vec4 FragColor2;
-layout (location = 3) out vec4 FragColor3;
-layout (location = 4) out vec4 FragColor4;
-layout (location = 5) out vec4 FragColor5;
-layout (location = 6) out vec4 FragColor6;
-layout (location = 7) out vec4 FragColor7;
-
-layout (std140, binding = {}) uniform fs_config {{
-    float y_direction;
-}};
-
-)",
-                       EmulationUniformBlockBinding);
-    out += Decompile(device, ir, ShaderType::Fragment, "fragment");
-
-    out += R"(
-void main() {
-    execute_fragment();
-}
-)";
-    return out;
-}
-
-std::string GenerateComputeShader(const Device& device, const ShaderIR& ir) {
-    std::string out = GetCommonDeclarations();
-    out += Decompile(device, ir, ShaderType::Compute, "compute");
-    out += R"(
-void main() {
-    execute_compute();
-}
-)";
-    return out;
-}
-
-} // 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
deleted file mode 100644
index cba2be9f9..000000000
--- a/src/video_core/renderer_opengl/gl_shader_gen.h
+++ /dev/null
@@ -1,34 +0,0 @@
-// Copyright 2018 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <vector>
-
-#include "common/common_types.h"
-#include "video_core/renderer_opengl/gl_shader_decompiler.h"
-#include "video_core/shader/shader_ir.h"
-
-namespace OpenGL {
-class Device;
-}
-
-namespace OpenGL::GLShader {
-
-using VideoCommon::Shader::ProgramCode;
-using VideoCommon::Shader::ShaderIR;
-
-/// Generates the GLSL vertex shader program source code for the given VS program
-std::string GenerateVertexShader(const Device& device, const ShaderIR& ir, const ShaderIR* ir_b);
-
-/// Generates the GLSL geometry shader program source code for the given GS program
-std::string GenerateGeometryShader(const Device& device, const ShaderIR& ir);
-
-/// Generates the GLSL fragment shader program source code for the given FS program
-std::string GenerateFragmentShader(const Device& device, const ShaderIR& ir);
-
-/// Generates the GLSL compute shader program source code for the given CS program
-std::string GenerateComputeShader(const Device& device, const ShaderIR& ir);
-
-} // namespace OpenGL::GLShader
diff --git a/src/video_core/renderer_opengl/gl_shader_manager.cpp b/src/video_core/renderer_opengl/gl_shader_manager.cpp
index 75d3fac04..9c7b0adbd 100644
--- a/src/video_core/renderer_opengl/gl_shader_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_manager.cpp
@@ -2,45 +2,52 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <glad/glad.h>
+
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_shader_manager.h"
 
 namespace OpenGL::GLShader {
 
-using Tegra::Engines::Maxwell3D;
-
-ProgramManager::ProgramManager() {
-    pipeline.Create();
-}
+ProgramManager::ProgramManager() = default;
 
 ProgramManager::~ProgramManager() = default;
 
-void ProgramManager::ApplyTo(OpenGLState& state) {
-    UpdatePipeline();
-    state.draw.shader_program = 0;
-    state.draw.program_pipeline = pipeline.handle;
+void ProgramManager::Create() {
+    graphics_pipeline.Create();
+    glBindProgramPipeline(graphics_pipeline.handle);
 }
 
-void ProgramManager::UpdatePipeline() {
+void ProgramManager::BindGraphicsPipeline() {
+    if (!is_graphics_bound) {
+        is_graphics_bound = true;
+        glUseProgram(0);
+    }
+
     // Avoid updating the pipeline when values have no changed
     if (old_state == current_state) {
         return;
     }
 
     // Workaround for AMD bug
-    constexpr GLenum all_used_stages{GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT |
-                                     GL_FRAGMENT_SHADER_BIT};
-    glUseProgramStages(pipeline.handle, all_used_stages, 0);
-
-    glUseProgramStages(pipeline.handle, GL_VERTEX_SHADER_BIT, current_state.vertex_shader);
-    glUseProgramStages(pipeline.handle, GL_GEOMETRY_SHADER_BIT, current_state.geometry_shader);
-    glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, current_state.fragment_shader);
+    static constexpr GLenum all_used_stages{GL_VERTEX_SHADER_BIT | GL_GEOMETRY_SHADER_BIT |
+                                            GL_FRAGMENT_SHADER_BIT};
+    const GLuint handle = graphics_pipeline.handle;
+    glUseProgramStages(handle, all_used_stages, 0);
+    glUseProgramStages(handle, GL_VERTEX_SHADER_BIT, current_state.vertex_shader);
+    glUseProgramStages(handle, GL_GEOMETRY_SHADER_BIT, current_state.geometry_shader);
+    glUseProgramStages(handle, GL_FRAGMENT_SHADER_BIT, current_state.fragment_shader);
 
     old_state = current_state;
 }
 
-void MaxwellUniformData::SetFromRegs(const Maxwell3D& maxwell) {
+void ProgramManager::BindComputeShader(GLuint program) {
+    is_graphics_bound = false;
+    glUseProgram(program);
+}
+
+void MaxwellUniformData::SetFromRegs(const Tegra::Engines::Maxwell3D& maxwell) {
     const auto& regs = maxwell.regs;
 
     // Y_NEGATE controls what value S2R returns for the Y_DIRECTION system value.
diff --git a/src/video_core/renderer_opengl/gl_shader_manager.h b/src/video_core/renderer_opengl/gl_shader_manager.h
index 478c165ce..d2e47f2a9 100644
--- a/src/video_core/renderer_opengl/gl_shader_manager.h
+++ b/src/video_core/renderer_opengl/gl_shader_manager.h
@@ -9,7 +9,6 @@
 #include <glad/glad.h>
 
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/maxwell_to_gl.h"
 
 namespace OpenGL::GLShader {
@@ -32,49 +31,47 @@ public:
     explicit ProgramManager();
     ~ProgramManager();
 
-    void ApplyTo(OpenGLState& state);
+    void Create();
 
-    void UseProgrammableVertexShader(GLuint program) {
+    /// Updates the graphics pipeline and binds it.
+    void BindGraphicsPipeline();
+
+    /// Binds a compute shader.
+    void BindComputeShader(GLuint program);
+
+    void UseVertexShader(GLuint program) {
         current_state.vertex_shader = program;
     }
 
-    void UseProgrammableGeometryShader(GLuint program) {
+    void UseGeometryShader(GLuint program) {
         current_state.geometry_shader = program;
     }
 
-    void UseProgrammableFragmentShader(GLuint program) {
+    void UseFragmentShader(GLuint program) {
         current_state.fragment_shader = program;
     }
 
-    void UseTrivialGeometryShader() {
-        current_state.geometry_shader = 0;
-    }
-
-    void UseTrivialFragmentShader() {
-        current_state.fragment_shader = 0;
-    }
-
 private:
     struct PipelineState {
-        bool operator==(const PipelineState& rhs) const {
+        bool operator==(const PipelineState& rhs) const noexcept {
             return vertex_shader == rhs.vertex_shader && fragment_shader == rhs.fragment_shader &&
                    geometry_shader == rhs.geometry_shader;
         }
 
-        bool operator!=(const PipelineState& rhs) const {
+        bool operator!=(const PipelineState& rhs) const noexcept {
             return !operator==(rhs);
         }
 
-        GLuint vertex_shader{};
-        GLuint fragment_shader{};
-        GLuint geometry_shader{};
+        GLuint vertex_shader = 0;
+        GLuint fragment_shader = 0;
+        GLuint geometry_shader = 0;
     };
 
-    void UpdatePipeline();
-
-    OGLPipeline pipeline;
+    OGLPipeline graphics_pipeline;
+    OGLPipeline compute_pipeline;
     PipelineState current_state;
     PipelineState old_state;
+    bool is_graphics_bound = true;
 };
 
 } // namespace OpenGL::GLShader
diff --git a/src/video_core/renderer_opengl/gl_state.cpp b/src/video_core/renderer_opengl/gl_state.cpp
deleted file mode 100644
index ab1f7983c..000000000
--- a/src/video_core/renderer_opengl/gl_state.cpp
+++ /dev/null
@@ -1,554 +0,0 @@
-// Copyright 2015 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <iterator>
-#include <glad/glad.h>
-#include "common/assert.h"
-#include "common/logging/log.h"
-#include "common/microprofile.h"
-#include "video_core/renderer_opengl/gl_state.h"
-
-MICROPROFILE_DEFINE(OpenGL_State, "OpenGL", "State Change", MP_RGB(192, 128, 128));
-
-namespace OpenGL {
-
-using Maxwell = Tegra::Engines::Maxwell3D::Regs;
-
-OpenGLState OpenGLState::cur_state;
-
-namespace {
-
-template <typename T>
-bool UpdateValue(T& current_value, const T new_value) {
-    const bool changed = current_value != new_value;
-    current_value = new_value;
-    return changed;
-}
-
-template <typename T1, typename T2>
-bool UpdateTie(T1 current_value, const T2 new_value) {
-    const bool changed = current_value != new_value;
-    current_value = new_value;
-    return changed;
-}
-
-template <typename T>
-std::optional<std::pair<GLuint, GLsizei>> UpdateArray(T& current_values, const T& new_values) {
-    std::optional<std::size_t> first;
-    std::size_t last;
-    for (std::size_t i = 0; i < std::size(current_values); ++i) {
-        if (!UpdateValue(current_values[i], new_values[i])) {
-            continue;
-        }
-        if (!first) {
-            first = i;
-        }
-        last = i;
-    }
-    if (!first) {
-        return std::nullopt;
-    }
-    return std::make_pair(static_cast<GLuint>(*first), static_cast<GLsizei>(last - *first + 1));
-}
-
-void Enable(GLenum cap, bool enable) {
-    if (enable) {
-        glEnable(cap);
-    } else {
-        glDisable(cap);
-    }
-}
-
-void Enable(GLenum cap, GLuint index, bool enable) {
-    if (enable) {
-        glEnablei(cap, index);
-    } else {
-        glDisablei(cap, index);
-    }
-}
-
-void Enable(GLenum cap, bool& current_value, bool new_value) {
-    if (UpdateValue(current_value, new_value)) {
-        Enable(cap, new_value);
-    }
-}
-
-void Enable(GLenum cap, GLuint index, bool& current_value, bool new_value) {
-    if (UpdateValue(current_value, new_value)) {
-        Enable(cap, index, new_value);
-    }
-}
-
-} // Anonymous namespace
-
-OpenGLState::OpenGLState() = default;
-
-void OpenGLState::SetDefaultViewports() {
-    viewports.fill(Viewport{});
-
-    depth_clamp.far_plane = false;
-    depth_clamp.near_plane = false;
-}
-
-void OpenGLState::ApplyFramebufferState() {
-    if (UpdateValue(cur_state.draw.read_framebuffer, draw.read_framebuffer)) {
-        glBindFramebuffer(GL_READ_FRAMEBUFFER, draw.read_framebuffer);
-    }
-    if (UpdateValue(cur_state.draw.draw_framebuffer, draw.draw_framebuffer)) {
-        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, draw.draw_framebuffer);
-    }
-}
-
-void OpenGLState::ApplyVertexArrayState() {
-    if (UpdateValue(cur_state.draw.vertex_array, draw.vertex_array)) {
-        glBindVertexArray(draw.vertex_array);
-    }
-}
-
-void OpenGLState::ApplyShaderProgram() {
-    if (UpdateValue(cur_state.draw.shader_program, draw.shader_program)) {
-        glUseProgram(draw.shader_program);
-    }
-}
-
-void OpenGLState::ApplyProgramPipeline() {
-    if (UpdateValue(cur_state.draw.program_pipeline, draw.program_pipeline)) {
-        glBindProgramPipeline(draw.program_pipeline);
-    }
-}
-
-void OpenGLState::ApplyClipDistances() {
-    for (std::size_t i = 0; i < clip_distance.size(); ++i) {
-        Enable(GL_CLIP_DISTANCE0 + static_cast<GLenum>(i), cur_state.clip_distance[i],
-               clip_distance[i]);
-    }
-}
-
-void OpenGLState::ApplyPointSize() {
-    Enable(GL_PROGRAM_POINT_SIZE, cur_state.point.program_control, point.program_control);
-    Enable(GL_POINT_SPRITE, cur_state.point.sprite, point.sprite);
-    if (UpdateValue(cur_state.point.size, point.size)) {
-        glPointSize(point.size);
-    }
-}
-
-void OpenGLState::ApplyFragmentColorClamp() {
-    if (UpdateValue(cur_state.fragment_color_clamp.enabled, fragment_color_clamp.enabled)) {
-        glClampColor(GL_CLAMP_FRAGMENT_COLOR_ARB,
-                     fragment_color_clamp.enabled ? GL_TRUE : GL_FALSE);
-    }
-}
-
-void OpenGLState::ApplyMultisample() {
-    Enable(GL_SAMPLE_ALPHA_TO_COVERAGE, cur_state.multisample_control.alpha_to_coverage,
-           multisample_control.alpha_to_coverage);
-    Enable(GL_SAMPLE_ALPHA_TO_ONE, cur_state.multisample_control.alpha_to_one,
-           multisample_control.alpha_to_one);
-}
-
-void OpenGLState::ApplyDepthClamp() {
-    if (depth_clamp.far_plane == cur_state.depth_clamp.far_plane &&
-        depth_clamp.near_plane == cur_state.depth_clamp.near_plane) {
-        return;
-    }
-    cur_state.depth_clamp = depth_clamp;
-
-    UNIMPLEMENTED_IF_MSG(depth_clamp.far_plane != depth_clamp.near_plane,
-                         "Unimplemented Depth Clamp Separation!");
-
-    Enable(GL_DEPTH_CLAMP, depth_clamp.far_plane || depth_clamp.near_plane);
-}
-
-void OpenGLState::ApplySRgb() {
-    if (cur_state.framebuffer_srgb.enabled == framebuffer_srgb.enabled)
-        return;
-    cur_state.framebuffer_srgb.enabled = framebuffer_srgb.enabled;
-    if (framebuffer_srgb.enabled) {
-        glEnable(GL_FRAMEBUFFER_SRGB);
-    } else {
-        glDisable(GL_FRAMEBUFFER_SRGB);
-    }
-}
-
-void OpenGLState::ApplyCulling() {
-    Enable(GL_CULL_FACE, cur_state.cull.enabled, cull.enabled);
-
-    if (UpdateValue(cur_state.cull.mode, cull.mode)) {
-        glCullFace(cull.mode);
-    }
-
-    if (UpdateValue(cur_state.cull.front_face, cull.front_face)) {
-        glFrontFace(cull.front_face);
-    }
-}
-
-void OpenGLState::ApplyRasterizerDiscard() {
-    Enable(GL_RASTERIZER_DISCARD, cur_state.rasterizer_discard, rasterizer_discard);
-}
-
-void OpenGLState::ApplyColorMask() {
-    if (!dirty.color_mask) {
-        return;
-    }
-    dirty.color_mask = false;
-
-    for (std::size_t i = 0; i < Maxwell::NumRenderTargets; ++i) {
-        const auto& updated = color_mask[i];
-        auto& current = cur_state.color_mask[i];
-        if (updated.red_enabled != current.red_enabled ||
-            updated.green_enabled != current.green_enabled ||
-            updated.blue_enabled != current.blue_enabled ||
-            updated.alpha_enabled != current.alpha_enabled) {
-            current = updated;
-            glColorMaski(static_cast<GLuint>(i), updated.red_enabled, updated.green_enabled,
-                         updated.blue_enabled, updated.alpha_enabled);
-        }
-    }
-}
-
-void OpenGLState::ApplyDepth() {
-    Enable(GL_DEPTH_TEST, cur_state.depth.test_enabled, depth.test_enabled);
-
-    if (cur_state.depth.test_func != depth.test_func) {
-        cur_state.depth.test_func = depth.test_func;
-        glDepthFunc(depth.test_func);
-    }
-
-    if (cur_state.depth.write_mask != depth.write_mask) {
-        cur_state.depth.write_mask = depth.write_mask;
-        glDepthMask(depth.write_mask);
-    }
-}
-
-void OpenGLState::ApplyPrimitiveRestart() {
-    Enable(GL_PRIMITIVE_RESTART, cur_state.primitive_restart.enabled, primitive_restart.enabled);
-
-    if (cur_state.primitive_restart.index != primitive_restart.index) {
-        cur_state.primitive_restart.index = primitive_restart.index;
-        glPrimitiveRestartIndex(primitive_restart.index);
-    }
-}
-
-void OpenGLState::ApplyStencilTest() {
-    if (!dirty.stencil_state) {
-        return;
-    }
-    dirty.stencil_state = false;
-
-    Enable(GL_STENCIL_TEST, cur_state.stencil.test_enabled, stencil.test_enabled);
-
-    const auto ConfigStencil = [](GLenum face, const auto& config, auto& current) {
-        if (current.test_func != config.test_func || current.test_ref != config.test_ref ||
-            current.test_mask != config.test_mask) {
-            current.test_func = config.test_func;
-            current.test_ref = config.test_ref;
-            current.test_mask = config.test_mask;
-            glStencilFuncSeparate(face, config.test_func, config.test_ref, config.test_mask);
-        }
-        if (current.action_depth_fail != config.action_depth_fail ||
-            current.action_depth_pass != config.action_depth_pass ||
-            current.action_stencil_fail != config.action_stencil_fail) {
-            current.action_depth_fail = config.action_depth_fail;
-            current.action_depth_pass = config.action_depth_pass;
-            current.action_stencil_fail = config.action_stencil_fail;
-            glStencilOpSeparate(face, config.action_stencil_fail, config.action_depth_fail,
-                                config.action_depth_pass);
-        }
-        if (current.write_mask != config.write_mask) {
-            current.write_mask = 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);
-}
-
-void OpenGLState::ApplyViewport() {
-    for (GLuint i = 0; i < static_cast<GLuint>(Maxwell::NumViewports); ++i) {
-        const auto& updated = viewports[i];
-        auto& current = cur_state.viewports[i];
-
-        if (current.x != updated.x || current.y != updated.y || current.width != updated.width ||
-            current.height != updated.height) {
-            current.x = updated.x;
-            current.y = updated.y;
-            current.width = updated.width;
-            current.height = updated.height;
-            glViewportIndexedf(i, static_cast<GLfloat>(updated.x), static_cast<GLfloat>(updated.y),
-                               static_cast<GLfloat>(updated.width),
-                               static_cast<GLfloat>(updated.height));
-        }
-        if (current.depth_range_near != updated.depth_range_near ||
-            current.depth_range_far != updated.depth_range_far) {
-            current.depth_range_near = updated.depth_range_near;
-            current.depth_range_far = updated.depth_range_far;
-            glDepthRangeIndexed(i, updated.depth_range_near, updated.depth_range_far);
-        }
-
-        Enable(GL_SCISSOR_TEST, i, current.scissor.enabled, updated.scissor.enabled);
-
-        if (current.scissor.x != updated.scissor.x || current.scissor.y != updated.scissor.y ||
-            current.scissor.width != updated.scissor.width ||
-            current.scissor.height != updated.scissor.height) {
-            current.scissor.x = updated.scissor.x;
-            current.scissor.y = updated.scissor.y;
-            current.scissor.width = updated.scissor.width;
-            current.scissor.height = updated.scissor.height;
-            glScissorIndexed(i, updated.scissor.x, updated.scissor.y, updated.scissor.width,
-                             updated.scissor.height);
-        }
-    }
-}
-
-void OpenGLState::ApplyGlobalBlending() {
-    const Blend& updated = blend[0];
-    Blend& current = cur_state.blend[0];
-
-    Enable(GL_BLEND, current.enabled, updated.enabled);
-
-    if (current.src_rgb_func != updated.src_rgb_func ||
-        current.dst_rgb_func != updated.dst_rgb_func || current.src_a_func != updated.src_a_func ||
-        current.dst_a_func != updated.dst_a_func) {
-        current.src_rgb_func = updated.src_rgb_func;
-        current.dst_rgb_func = updated.dst_rgb_func;
-        current.src_a_func = updated.src_a_func;
-        current.dst_a_func = updated.dst_a_func;
-        glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
-                            updated.dst_a_func);
-    }
-
-    if (current.rgb_equation != updated.rgb_equation || current.a_equation != updated.a_equation) {
-        current.rgb_equation = updated.rgb_equation;
-        current.a_equation = updated.a_equation;
-        glBlendEquationSeparate(updated.rgb_equation, updated.a_equation);
-    }
-}
-
-void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) {
-    const Blend& updated = blend[target];
-    Blend& current = cur_state.blend[target];
-
-    if (current.enabled != updated.enabled || force) {
-        current.enabled = updated.enabled;
-        Enable(GL_BLEND, static_cast<GLuint>(target), updated.enabled);
-    }
-
-    if (UpdateTie(std::tie(current.src_rgb_func, current.dst_rgb_func, current.src_a_func,
-                           current.dst_a_func),
-                  std::tie(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
-                           updated.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 (UpdateTie(std::tie(current.rgb_equation, current.a_equation),
-                  std::tie(updated.rgb_equation, updated.a_equation))) {
-        glBlendEquationSeparatei(static_cast<GLuint>(target), updated.rgb_equation,
-                                 updated.a_equation);
-    }
-}
-
-void OpenGLState::ApplyBlending() {
-    if (!dirty.blend_state) {
-        return;
-    }
-    dirty.blend_state = false;
-
-    if (independant_blend.enabled) {
-        const bool force = independant_blend.enabled != cur_state.independant_blend.enabled;
-        for (std::size_t target = 0; target < Maxwell::NumRenderTargets; ++target) {
-            ApplyTargetBlending(target, force);
-        }
-    } else {
-        ApplyGlobalBlending();
-    }
-    cur_state.independant_blend.enabled = independant_blend.enabled;
-
-    if (UpdateTie(
-            std::tie(cur_state.blend_color.red, cur_state.blend_color.green,
-                     cur_state.blend_color.blue, cur_state.blend_color.alpha),
-            std::tie(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha))) {
-        glBlendColor(blend_color.red, blend_color.green, blend_color.blue, blend_color.alpha);
-    }
-}
-
-void OpenGLState::ApplyLogicOp() {
-    Enable(GL_COLOR_LOGIC_OP, cur_state.logic_op.enabled, logic_op.enabled);
-
-    if (UpdateValue(cur_state.logic_op.operation, logic_op.operation)) {
-        glLogicOp(logic_op.operation);
-    }
-}
-
-void OpenGLState::ApplyPolygonOffset() {
-    if (!dirty.polygon_offset) {
-        return;
-    }
-    dirty.polygon_offset = false;
-
-    Enable(GL_POLYGON_OFFSET_FILL, cur_state.polygon_offset.fill_enable,
-           polygon_offset.fill_enable);
-    Enable(GL_POLYGON_OFFSET_LINE, cur_state.polygon_offset.line_enable,
-           polygon_offset.line_enable);
-    Enable(GL_POLYGON_OFFSET_POINT, cur_state.polygon_offset.point_enable,
-           polygon_offset.point_enable);
-
-    if (UpdateTie(std::tie(cur_state.polygon_offset.factor, cur_state.polygon_offset.units,
-                           cur_state.polygon_offset.clamp),
-                  std::tie(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp))) {
-        if (GLAD_GL_EXT_polygon_offset_clamp && polygon_offset.clamp != 0) {
-            glPolygonOffsetClamp(polygon_offset.factor, polygon_offset.units, polygon_offset.clamp);
-        } else {
-            UNIMPLEMENTED_IF_MSG(polygon_offset.clamp != 0,
-                                 "Unimplemented Depth polygon offset clamp.");
-            glPolygonOffset(polygon_offset.factor, polygon_offset.units);
-        }
-    }
-}
-
-void OpenGLState::ApplyAlphaTest() {
-    Enable(GL_ALPHA_TEST, cur_state.alpha_test.enabled, alpha_test.enabled);
-    if (UpdateTie(std::tie(cur_state.alpha_test.func, cur_state.alpha_test.ref),
-                  std::tie(alpha_test.func, alpha_test.ref))) {
-        glAlphaFunc(alpha_test.func, alpha_test.ref);
-    }
-}
-
-void OpenGLState::ApplyClipControl() {
-    if (UpdateTie(std::tie(cur_state.clip_control.origin, cur_state.clip_control.depth_mode),
-                  std::tie(clip_control.origin, clip_control.depth_mode))) {
-        glClipControl(clip_control.origin, clip_control.depth_mode);
-    }
-}
-
-void OpenGLState::ApplyTextures() {
-    const std::size_t size = std::size(textures);
-    for (std::size_t i = 0; i < size; ++i) {
-        if (UpdateValue(cur_state.textures[i], textures[i])) {
-            // BindTextureUnit doesn't support binding null textures, skip those binds.
-            // TODO(Rodrigo): Stop using null textures
-            if (textures[i] != 0) {
-                glBindTextureUnit(static_cast<GLuint>(i), textures[i]);
-            }
-        }
-    }
-}
-
-void OpenGLState::ApplySamplers() {
-    const std::size_t size = std::size(samplers);
-    for (std::size_t i = 0; i < size; ++i) {
-        if (UpdateValue(cur_state.samplers[i], samplers[i])) {
-            glBindSampler(static_cast<GLuint>(i), samplers[i]);
-        }
-    }
-}
-
-void OpenGLState::ApplyImages() {
-    if (const auto update = UpdateArray(cur_state.images, images)) {
-        glBindImageTextures(update->first, update->second, images.data() + update->first);
-    }
-}
-
-void OpenGLState::Apply() {
-    MICROPROFILE_SCOPE(OpenGL_State);
-    ApplyFramebufferState();
-    ApplyVertexArrayState();
-    ApplyShaderProgram();
-    ApplyProgramPipeline();
-    ApplyClipDistances();
-    ApplyPointSize();
-    ApplyFragmentColorClamp();
-    ApplyMultisample();
-    ApplyRasterizerDiscard();
-    ApplyColorMask();
-    ApplyDepthClamp();
-    ApplyViewport();
-    ApplyStencilTest();
-    ApplySRgb();
-    ApplyCulling();
-    ApplyDepth();
-    ApplyPrimitiveRestart();
-    ApplyBlending();
-    ApplyLogicOp();
-    ApplyTextures();
-    ApplySamplers();
-    ApplyImages();
-    ApplyPolygonOffset();
-    ApplyAlphaTest();
-    ApplyClipControl();
-}
-
-void OpenGLState::EmulateViewportWithScissor() {
-    auto& current = viewports[0];
-    if (current.scissor.enabled) {
-        const GLint left = std::max(current.x, current.scissor.x);
-        const GLint right =
-            std::max(current.x + current.width, current.scissor.x + current.scissor.width);
-        const GLint bottom = std::max(current.y, current.scissor.y);
-        const GLint top =
-            std::max(current.y + current.height, current.scissor.y + current.scissor.height);
-        current.scissor.x = std::max(left, 0);
-        current.scissor.y = std::max(bottom, 0);
-        current.scissor.width = std::max(right - left, 0);
-        current.scissor.height = std::max(top - bottom, 0);
-    } else {
-        current.scissor.enabled = true;
-        current.scissor.x = current.x;
-        current.scissor.y = current.y;
-        current.scissor.width = current.width;
-        current.scissor.height = current.height;
-    }
-}
-
-OpenGLState& OpenGLState::UnbindTexture(GLuint handle) {
-    for (auto& texture : textures) {
-        if (texture == handle) {
-            texture = 0;
-        }
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetSampler(GLuint handle) {
-    for (auto& sampler : samplers) {
-        if (sampler == handle) {
-            sampler = 0;
-        }
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetProgram(GLuint handle) {
-    if (draw.shader_program == handle) {
-        draw.shader_program = 0;
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetPipeline(GLuint handle) {
-    if (draw.program_pipeline == handle) {
-        draw.program_pipeline = 0;
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetVertexArray(GLuint handle) {
-    if (draw.vertex_array == handle) {
-        draw.vertex_array = 0;
-    }
-    return *this;
-}
-
-OpenGLState& OpenGLState::ResetFramebuffer(GLuint handle) {
-    if (draw.read_framebuffer == handle) {
-        draw.read_framebuffer = 0;
-    }
-    if (draw.draw_framebuffer == handle) {
-        draw.draw_framebuffer = 0;
-    }
-    return *this;
-}
-
-} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_state.h b/src/video_core/renderer_opengl/gl_state.h
deleted file mode 100644
index 4953eeda2..000000000
--- a/src/video_core/renderer_opengl/gl_state.h
+++ /dev/null
@@ -1,247 +0,0 @@
-// Copyright 2015 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <type_traits>
-#include <glad/glad.h>
-#include "video_core/engines/maxwell_3d.h"
-
-namespace OpenGL {
-
-class OpenGLState {
-public:
-    struct {
-        bool enabled = false; // GL_FRAMEBUFFER_SRGB
-    } framebuffer_srgb;
-
-    struct {
-        bool alpha_to_coverage = false; // GL_ALPHA_TO_COVERAGE
-        bool alpha_to_one = false;      // GL_ALPHA_TO_ONE
-    } multisample_control;
-
-    struct {
-        bool enabled = false; // GL_CLAMP_FRAGMENT_COLOR_ARB
-    } fragment_color_clamp;
-
-    struct {
-        bool far_plane = false;
-        bool near_plane = false;
-    } depth_clamp; // GL_DEPTH_CLAMP
-
-    struct {
-        bool enabled = false;       // GL_CULL_FACE
-        GLenum mode = GL_BACK;      // GL_CULL_FACE_MODE
-        GLenum front_face = GL_CCW; // GL_FRONT_FACE
-    } cull;
-
-    struct {
-        bool test_enabled = false;      // GL_DEPTH_TEST
-        GLboolean write_mask = GL_TRUE; // GL_DEPTH_WRITEMASK
-        GLenum test_func = GL_LESS;     // GL_DEPTH_FUNC
-    } depth;
-
-    struct {
-        bool enabled = false;
-        GLuint index = 0;
-    } primitive_restart; // GL_PRIMITIVE_RESTART
-
-    bool rasterizer_discard = false; // GL_RASTERIZER_DISCARD
-
-    struct ColorMask {
-        GLboolean red_enabled = GL_TRUE;
-        GLboolean green_enabled = GL_TRUE;
-        GLboolean blue_enabled = GL_TRUE;
-        GLboolean alpha_enabled = GL_TRUE;
-    };
-    std::array<ColorMask, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets>
-        color_mask; // GL_COLOR_WRITEMASK
-
-    struct {
-        bool test_enabled = false; // GL_STENCIL_TEST
-        struct {
-            GLenum test_func = GL_ALWAYS;         // GL_STENCIL_FUNC
-            GLint test_ref = 0;                   // GL_STENCIL_REF
-            GLuint test_mask = 0xFFFFFFFF;        // GL_STENCIL_VALUE_MASK
-            GLuint write_mask = 0xFFFFFFFF;       // GL_STENCIL_WRITEMASK
-            GLenum action_stencil_fail = GL_KEEP; // GL_STENCIL_FAIL
-            GLenum action_depth_fail = GL_KEEP;   // GL_STENCIL_PASS_DEPTH_FAIL
-            GLenum action_depth_pass = GL_KEEP;   // GL_STENCIL_PASS_DEPTH_PASS
-        } front, back;
-    } stencil;
-
-    struct Blend {
-        bool enabled = false;              // GL_BLEND
-        GLenum rgb_equation = GL_FUNC_ADD; // GL_BLEND_EQUATION_RGB
-        GLenum a_equation = GL_FUNC_ADD;   // GL_BLEND_EQUATION_ALPHA
-        GLenum src_rgb_func = GL_ONE;      // GL_BLEND_SRC_RGB
-        GLenum dst_rgb_func = GL_ZERO;     // GL_BLEND_DST_RGB
-        GLenum src_a_func = GL_ONE;        // GL_BLEND_SRC_ALPHA
-        GLenum dst_a_func = GL_ZERO;       // GL_BLEND_DST_ALPHA
-    };
-    std::array<Blend, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> blend;
-
-    struct {
-        bool enabled = false;
-    } independant_blend;
-
-    struct {
-        GLclampf red = 0.0f;
-        GLclampf green = 0.0f;
-        GLclampf blue = 0.0f;
-        GLclampf alpha = 0.0f;
-    } blend_color; // GL_BLEND_COLOR
-
-    struct {
-        bool enabled = false; // GL_LOGIC_OP_MODE
-        GLenum operation = GL_COPY;
-    } logic_op;
-
-    static constexpr std::size_t NumSamplers = 32 * 5;
-    static constexpr std::size_t NumImages = 8 * 5;
-    std::array<GLuint, NumSamplers> textures = {};
-    std::array<GLuint, NumSamplers> samplers = {};
-    std::array<GLuint, NumImages> images = {};
-
-    struct {
-        GLuint read_framebuffer = 0; // GL_READ_FRAMEBUFFER_BINDING
-        GLuint draw_framebuffer = 0; // GL_DRAW_FRAMEBUFFER_BINDING
-        GLuint vertex_array = 0;     // GL_VERTEX_ARRAY_BINDING
-        GLuint shader_program = 0;   // GL_CURRENT_PROGRAM
-        GLuint program_pipeline = 0; // GL_PROGRAM_PIPELINE_BINDING
-    } draw;
-
-    struct Viewport {
-        GLint x = 0;
-        GLint y = 0;
-        GLint width = 0;
-        GLint height = 0;
-        GLfloat depth_range_near = 0.0f; // GL_DEPTH_RANGE
-        GLfloat depth_range_far = 1.0f;  // GL_DEPTH_RANGE
-        struct {
-            bool enabled = false; // GL_SCISSOR_TEST
-            GLint x = 0;
-            GLint y = 0;
-            GLsizei width = 0;
-            GLsizei height = 0;
-        } scissor;
-    };
-    std::array<Viewport, Tegra::Engines::Maxwell3D::Regs::NumViewports> viewports;
-
-    struct {
-        bool program_control = false; // GL_PROGRAM_POINT_SIZE
-        bool sprite = false;          // GL_POINT_SPRITE
-        GLfloat size = 1.0f;          // GL_POINT_SIZE
-    } point;
-
-    struct {
-        bool point_enable = false;
-        bool line_enable = false;
-        bool fill_enable = false;
-        GLfloat units = 0.0f;
-        GLfloat factor = 0.0f;
-        GLfloat clamp = 0.0f;
-    } polygon_offset;
-
-    struct {
-        bool enabled = false;    // GL_ALPHA_TEST
-        GLenum func = GL_ALWAYS; // GL_ALPHA_TEST_FUNC
-        GLfloat ref = 0.0f;      // GL_ALPHA_TEST_REF
-    } alpha_test;
-
-    std::array<bool, 8> clip_distance = {}; // GL_CLIP_DISTANCE
-
-    struct {
-        GLenum origin = GL_LOWER_LEFT;
-        GLenum depth_mode = GL_NEGATIVE_ONE_TO_ONE;
-    } clip_control;
-
-    OpenGLState();
-
-    /// Get the currently active OpenGL state
-    static OpenGLState GetCurState() {
-        return cur_state;
-    }
-
-    void SetDefaultViewports();
-    /// Apply this state as the current OpenGL state
-    void Apply();
-
-    void ApplyFramebufferState();
-    void ApplyVertexArrayState();
-    void ApplyShaderProgram();
-    void ApplyProgramPipeline();
-    void ApplyClipDistances();
-    void ApplyPointSize();
-    void ApplyFragmentColorClamp();
-    void ApplyMultisample();
-    void ApplySRgb();
-    void ApplyCulling();
-    void ApplyRasterizerDiscard();
-    void ApplyColorMask();
-    void ApplyDepth();
-    void ApplyPrimitiveRestart();
-    void ApplyStencilTest();
-    void ApplyViewport();
-    void ApplyTargetBlending(std::size_t target, bool force);
-    void ApplyGlobalBlending();
-    void ApplyBlending();
-    void ApplyLogicOp();
-    void ApplyTextures();
-    void ApplySamplers();
-    void ApplyImages();
-    void ApplyDepthClamp();
-    void ApplyPolygonOffset();
-    void ApplyAlphaTest();
-    void ApplyClipControl();
-
-    /// Resets any references to the given resource
-    OpenGLState& UnbindTexture(GLuint handle);
-    OpenGLState& ResetSampler(GLuint handle);
-    OpenGLState& ResetProgram(GLuint handle);
-    OpenGLState& ResetPipeline(GLuint handle);
-    OpenGLState& ResetVertexArray(GLuint handle);
-    OpenGLState& ResetFramebuffer(GLuint handle);
-
-    /// Viewport does not affects glClearBuffer so emulate viewport using scissor test
-    void EmulateViewportWithScissor();
-
-    void MarkDirtyBlendState() {
-        dirty.blend_state = true;
-    }
-
-    void MarkDirtyStencilState() {
-        dirty.stencil_state = true;
-    }
-
-    void MarkDirtyPolygonOffset() {
-        dirty.polygon_offset = true;
-    }
-
-    void MarkDirtyColorMask() {
-        dirty.color_mask = true;
-    }
-
-    void AllDirty() {
-        dirty.blend_state = true;
-        dirty.stencil_state = true;
-        dirty.polygon_offset = true;
-        dirty.color_mask = true;
-    }
-
-private:
-    static OpenGLState cur_state;
-
-    struct {
-        bool blend_state;
-        bool stencil_state;
-        bool viewport_state;
-        bool polygon_offset;
-        bool color_mask;
-    } dirty{};
-};
-static_assert(std::is_trivially_copyable_v<OpenGLState>);
-
-} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_state_tracker.cpp b/src/video_core/renderer_opengl/gl_state_tracker.cpp
new file mode 100644
index 000000000..255ac3147
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_state_tracker.cpp
@@ -0,0 +1,247 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <cstddef>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
+
+#define OFF(field_name) MAXWELL3D_REG_INDEX(field_name)
+#define NUM(field_name) (sizeof(Maxwell3D::Regs::field_name) / sizeof(u32))
+
+namespace OpenGL {
+
+namespace {
+
+using namespace Dirty;
+using namespace VideoCommon::Dirty;
+using Tegra::Engines::Maxwell3D;
+using Regs = Maxwell3D::Regs;
+using Tables = Maxwell3D::DirtyState::Tables;
+using Table = Maxwell3D::DirtyState::Table;
+
+void SetupDirtyColorMasks(Tables& tables) {
+    tables[0][OFF(color_mask_common)] = ColorMaskCommon;
+    for (std::size_t rt = 0; rt < Regs::NumRenderTargets; ++rt) {
+        const std::size_t offset = OFF(color_mask) + rt * NUM(color_mask[0]);
+        FillBlock(tables[0], offset, NUM(color_mask[0]), ColorMask0 + rt);
+    }
+
+    FillBlock(tables[1], OFF(color_mask), NUM(color_mask), ColorMasks);
+}
+
+void SetupDirtyVertexArrays(Tables& tables) {
+    static constexpr std::size_t num_array = 3;
+    static constexpr std::size_t instance_base_offset = 3;
+    for (std::size_t i = 0; i < Regs::NumVertexArrays; ++i) {
+        const std::size_t array_offset = OFF(vertex_array) + i * NUM(vertex_array[0]);
+        const std::size_t limit_offset = OFF(vertex_array_limit) + i * NUM(vertex_array_limit[0]);
+
+        FillBlock(tables, array_offset, num_array, VertexBuffer0 + i, VertexBuffers);
+        FillBlock(tables, limit_offset, NUM(vertex_array_limit), VertexBuffer0 + i, VertexBuffers);
+
+        const std::size_t instance_array_offset = array_offset + instance_base_offset;
+        tables[0][instance_array_offset] = static_cast<u8>(VertexInstance0 + i);
+        tables[1][instance_array_offset] = VertexInstances;
+
+        const std::size_t instance_offset = OFF(instanced_arrays) + i;
+        tables[0][instance_offset] = static_cast<u8>(VertexInstance0 + i);
+        tables[1][instance_offset] = VertexInstances;
+    }
+}
+
+void SetupDirtyVertexFormat(Tables& tables) {
+    for (std::size_t i = 0; i < Regs::NumVertexAttributes; ++i) {
+        const std::size_t offset = OFF(vertex_attrib_format) + i * NUM(vertex_attrib_format[0]);
+        FillBlock(tables[0], offset, NUM(vertex_attrib_format[0]), VertexFormat0 + i);
+    }
+
+    FillBlock(tables[1], OFF(vertex_attrib_format), Regs::NumVertexAttributes, VertexFormats);
+}
+
+void SetupDirtyViewports(Tables& tables) {
+    for (std::size_t i = 0; i < Regs::NumViewports; ++i) {
+        const std::size_t transf_offset = OFF(viewport_transform) + i * NUM(viewport_transform[0]);
+        const std::size_t viewport_offset = OFF(viewports) + i * NUM(viewports[0]);
+
+        FillBlock(tables[0], transf_offset, NUM(viewport_transform[0]), Viewport0 + i);
+        FillBlock(tables[0], viewport_offset, NUM(viewports[0]), Viewport0 + i);
+    }
+
+    FillBlock(tables[1], OFF(viewport_transform), NUM(viewport_transform), Viewports);
+    FillBlock(tables[1], OFF(viewports), NUM(viewports), Viewports);
+
+    tables[0][OFF(viewport_transform_enabled)] = ViewportTransform;
+    tables[1][OFF(viewport_transform_enabled)] = Viewports;
+}
+
+void SetupDirtyScissors(Tables& tables) {
+    for (std::size_t i = 0; i < Regs::NumViewports; ++i) {
+        const std::size_t offset = OFF(scissor_test) + i * NUM(scissor_test[0]);
+        FillBlock(tables[0], offset, NUM(scissor_test[0]), Scissor0 + i);
+    }
+    FillBlock(tables[1], OFF(scissor_test), NUM(scissor_test), Scissors);
+}
+
+void SetupDirtyShaders(Tables& tables) {
+    FillBlock(tables[0], OFF(shader_config[0]), NUM(shader_config[0]) * Regs::MaxShaderProgram,
+              Shaders);
+}
+
+void SetupDirtyPolygonModes(Tables& tables) {
+    tables[0][OFF(polygon_mode_front)] = PolygonModeFront;
+    tables[0][OFF(polygon_mode_back)] = PolygonModeBack;
+
+    tables[1][OFF(polygon_mode_front)] = PolygonModes;
+    tables[1][OFF(polygon_mode_back)] = PolygonModes;
+    tables[0][OFF(fill_rectangle)] = PolygonModes;
+}
+
+void SetupDirtyDepthTest(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(depth_test_enable)] = DepthTest;
+    table[OFF(depth_write_enabled)] = DepthMask;
+    table[OFF(depth_test_func)] = DepthTest;
+}
+
+void SetupDirtyStencilTest(Tables& tables) {
+    static constexpr std::array offsets = {
+        OFF(stencil_enable),          OFF(stencil_front_func_func), OFF(stencil_front_func_ref),
+        OFF(stencil_front_func_mask), OFF(stencil_front_op_fail),   OFF(stencil_front_op_zfail),
+        OFF(stencil_front_op_zpass),  OFF(stencil_front_mask),      OFF(stencil_two_side_enable),
+        OFF(stencil_back_func_func),  OFF(stencil_back_func_ref),   OFF(stencil_back_func_mask),
+        OFF(stencil_back_op_fail),    OFF(stencil_back_op_zfail),   OFF(stencil_back_op_zpass),
+        OFF(stencil_back_mask)};
+    for (const auto offset : offsets) {
+        tables[0][offset] = StencilTest;
+    }
+}
+
+void SetupDirtyAlphaTest(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(alpha_test_ref)] = AlphaTest;
+    table[OFF(alpha_test_func)] = AlphaTest;
+    table[OFF(alpha_test_enabled)] = AlphaTest;
+}
+
+void SetupDirtyBlend(Tables& tables) {
+    FillBlock(tables[0], OFF(blend_color), NUM(blend_color), BlendColor);
+
+    tables[0][OFF(independent_blend_enable)] = BlendIndependentEnabled;
+
+    for (std::size_t i = 0; i < Regs::NumRenderTargets; ++i) {
+        const std::size_t offset = OFF(independent_blend) + i * NUM(independent_blend[0]);
+        FillBlock(tables[0], offset, NUM(independent_blend[0]), BlendState0 + i);
+
+        tables[0][OFF(blend.enable) + i] = static_cast<u8>(BlendState0 + i);
+    }
+    FillBlock(tables[1], OFF(independent_blend), NUM(independent_blend), BlendStates);
+    FillBlock(tables[1], OFF(blend), NUM(blend), BlendStates);
+}
+
+void SetupDirtyPrimitiveRestart(Tables& tables) {
+    FillBlock(tables[0], OFF(primitive_restart), NUM(primitive_restart), PrimitiveRestart);
+}
+
+void SetupDirtyPolygonOffset(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(polygon_offset_fill_enable)] = PolygonOffset;
+    table[OFF(polygon_offset_line_enable)] = PolygonOffset;
+    table[OFF(polygon_offset_point_enable)] = PolygonOffset;
+    table[OFF(polygon_offset_factor)] = PolygonOffset;
+    table[OFF(polygon_offset_units)] = PolygonOffset;
+    table[OFF(polygon_offset_clamp)] = PolygonOffset;
+}
+
+void SetupDirtyMultisampleControl(Tables& tables) {
+    FillBlock(tables[0], OFF(multisample_control), NUM(multisample_control), MultisampleControl);
+}
+
+void SetupDirtyRasterizeEnable(Tables& tables) {
+    tables[0][OFF(rasterize_enable)] = RasterizeEnable;
+}
+
+void SetupDirtyFramebufferSRGB(Tables& tables) {
+    tables[0][OFF(framebuffer_srgb)] = FramebufferSRGB;
+}
+
+void SetupDirtyLogicOp(Tables& tables) {
+    FillBlock(tables[0], OFF(logic_op), NUM(logic_op), LogicOp);
+}
+
+void SetupDirtyFragmentClampColor(Tables& tables) {
+    tables[0][OFF(frag_color_clamp)] = FragmentClampColor;
+}
+
+void SetupDirtyPointSize(Tables& tables) {
+    tables[0][OFF(vp_point_size)] = PointSize;
+    tables[0][OFF(point_size)] = PointSize;
+    tables[0][OFF(point_sprite_enable)] = PointSize;
+}
+
+void SetupDirtyClipControl(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(screen_y_control)] = ClipControl;
+    table[OFF(depth_mode)] = ClipControl;
+}
+
+void SetupDirtyDepthClampEnabled(Tables& tables) {
+    tables[0][OFF(view_volume_clip_control)] = DepthClampEnabled;
+}
+
+void SetupDirtyMisc(Tables& tables) {
+    auto& table = tables[0];
+
+    table[OFF(clip_distance_enabled)] = ClipDistances;
+
+    table[OFF(front_face)] = FrontFace;
+
+    table[OFF(cull_test_enabled)] = CullTest;
+    table[OFF(cull_face)] = CullTest;
+}
+
+} // Anonymous namespace
+
+StateTracker::StateTracker(Core::System& system) : system{system} {}
+
+void StateTracker::Initialize() {
+    auto& dirty = system.GPU().Maxwell3D().dirty;
+    auto& tables = dirty.tables;
+    SetupDirtyRenderTargets(tables);
+    SetupDirtyColorMasks(tables);
+    SetupDirtyViewports(tables);
+    SetupDirtyScissors(tables);
+    SetupDirtyVertexArrays(tables);
+    SetupDirtyVertexFormat(tables);
+    SetupDirtyShaders(tables);
+    SetupDirtyPolygonModes(tables);
+    SetupDirtyDepthTest(tables);
+    SetupDirtyStencilTest(tables);
+    SetupDirtyAlphaTest(tables);
+    SetupDirtyBlend(tables);
+    SetupDirtyPrimitiveRestart(tables);
+    SetupDirtyPolygonOffset(tables);
+    SetupDirtyMultisampleControl(tables);
+    SetupDirtyRasterizeEnable(tables);
+    SetupDirtyFramebufferSRGB(tables);
+    SetupDirtyLogicOp(tables);
+    SetupDirtyFragmentClampColor(tables);
+    SetupDirtyPointSize(tables);
+    SetupDirtyClipControl(tables);
+    SetupDirtyDepthClampEnabled(tables);
+    SetupDirtyMisc(tables);
+
+    auto& store = dirty.on_write_stores;
+    store[VertexBuffers] = true;
+    for (std::size_t i = 0; i < Regs::NumVertexArrays; ++i) {
+        store[VertexBuffer0 + i] = true;
+    }
+}
+
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_state_tracker.h b/src/video_core/renderer_opengl/gl_state_tracker.h
new file mode 100644
index 000000000..b882d75c3
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_state_tracker.h
@@ -0,0 +1,215 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <limits>
+
+#include <glad/glad.h>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/dirty_flags.h"
+#include "video_core/engines/maxwell_3d.h"
+
+namespace Core {
+class System;
+}
+
+namespace OpenGL {
+
+namespace Dirty {
+
+enum : u8 {
+    First = VideoCommon::Dirty::LastCommonEntry,
+
+    VertexFormats,
+    VertexFormat0,
+    VertexFormat31 = VertexFormat0 + 31,
+
+    VertexBuffers,
+    VertexBuffer0,
+    VertexBuffer31 = VertexBuffer0 + 31,
+
+    VertexInstances,
+    VertexInstance0,
+    VertexInstance31 = VertexInstance0 + 31,
+
+    ViewportTransform,
+    Viewports,
+    Viewport0,
+    Viewport15 = Viewport0 + 15,
+
+    Scissors,
+    Scissor0,
+    Scissor15 = Scissor0 + 15,
+
+    ColorMaskCommon,
+    ColorMasks,
+    ColorMask0,
+    ColorMask7 = ColorMask0 + 7,
+
+    BlendColor,
+    BlendIndependentEnabled,
+    BlendStates,
+    BlendState0,
+    BlendState7 = BlendState0 + 7,
+
+    Shaders,
+    ClipDistances,
+
+    PolygonModes,
+    PolygonModeFront,
+    PolygonModeBack,
+
+    ColorMask,
+    FrontFace,
+    CullTest,
+    DepthMask,
+    DepthTest,
+    StencilTest,
+    AlphaTest,
+    PrimitiveRestart,
+    PolygonOffset,
+    MultisampleControl,
+    RasterizeEnable,
+    FramebufferSRGB,
+    LogicOp,
+    FragmentClampColor,
+    PointSize,
+    ClipControl,
+    DepthClampEnabled,
+
+    Last
+};
+static_assert(Last <= std::numeric_limits<u8>::max());
+
+} // namespace Dirty
+
+class StateTracker {
+public:
+    explicit StateTracker(Core::System& system);
+
+    void Initialize();
+
+    void BindIndexBuffer(GLuint new_index_buffer) {
+        if (index_buffer == new_index_buffer) {
+            return;
+        }
+        index_buffer = new_index_buffer;
+        glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, new_index_buffer);
+    }
+
+    void NotifyScreenDrawVertexArray() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::VertexFormats] = true;
+        flags[OpenGL::Dirty::VertexFormat0 + 0] = true;
+        flags[OpenGL::Dirty::VertexFormat0 + 1] = true;
+
+        flags[OpenGL::Dirty::VertexBuffers] = true;
+        flags[OpenGL::Dirty::VertexBuffer0] = true;
+
+        flags[OpenGL::Dirty::VertexInstances] = true;
+        flags[OpenGL::Dirty::VertexInstance0 + 0] = true;
+        flags[OpenGL::Dirty::VertexInstance0 + 1] = true;
+    }
+
+    void NotifyPolygonModes() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::PolygonModes] = true;
+        flags[OpenGL::Dirty::PolygonModeFront] = true;
+        flags[OpenGL::Dirty::PolygonModeBack] = true;
+    }
+
+    void NotifyViewport0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::Viewports] = true;
+        flags[OpenGL::Dirty::Viewport0] = true;
+    }
+
+    void NotifyScissor0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::Scissors] = true;
+        flags[OpenGL::Dirty::Scissor0] = true;
+    }
+
+    void NotifyColorMask0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::ColorMasks] = true;
+        flags[OpenGL::Dirty::ColorMask0] = true;
+    }
+
+    void NotifyBlend0() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::BlendStates] = true;
+        flags[OpenGL::Dirty::BlendState0] = true;
+    }
+
+    void NotifyFramebuffer() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[VideoCommon::Dirty::RenderTargets] = true;
+    }
+
+    void NotifyFrontFace() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::FrontFace] = true;
+    }
+
+    void NotifyCullTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::CullTest] = true;
+    }
+
+    void NotifyDepthMask() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::DepthMask] = true;
+    }
+
+    void NotifyDepthTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::DepthTest] = true;
+    }
+
+    void NotifyStencilTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::StencilTest] = true;
+    }
+
+    void NotifyPolygonOffset() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::PolygonOffset] = true;
+    }
+
+    void NotifyRasterizeEnable() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::RasterizeEnable] = true;
+    }
+
+    void NotifyFramebufferSRGB() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::FramebufferSRGB] = true;
+    }
+
+    void NotifyLogicOp() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::LogicOp] = true;
+    }
+
+    void NotifyClipControl() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::ClipControl] = true;
+    }
+
+    void NotifyAlphaTest() {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        flags[OpenGL::Dirty::AlphaTest] = true;
+    }
+
+private:
+    Core::System& system;
+
+    GLuint index_buffer = 0;
+};
+
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_stream_buffer.cpp b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
index 35ba334e4..6ec328c53 100644
--- a/src/video_core/renderer_opengl/gl_stream_buffer.cpp
+++ b/src/video_core/renderer_opengl/gl_stream_buffer.cpp
@@ -7,7 +7,6 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/microprofile.h"
-#include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_stream_buffer.h"
 
 MICROPROFILE_DEFINE(OpenGL_StreamBuffer, "OpenGL", "Stream Buffer Orphaning",
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index 5c1ae1418..f424e3000 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -10,7 +10,7 @@
 #include "core/core.h"
 #include "video_core/morton.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 #include "video_core/renderer_opengl/utils.h"
 #include "video_core/texture_cache/surface_base.h"
@@ -53,6 +53,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format
     {GL_R8UI, GL_RED_INTEGER, GL_UNSIGNED_BYTE, false},                             // R8UI
     {GL_RGBA16F, GL_RGBA, GL_HALF_FLOAT, false},                                    // RGBA16F
     {GL_RGBA16, GL_RGBA, GL_UNSIGNED_SHORT, false},                                 // RGBA16U
+    {GL_RGBA16_SNORM, GL_RGBA, GL_SHORT, false},                                    // RGBA16S
     {GL_RGBA16UI, GL_RGBA_INTEGER, GL_UNSIGNED_SHORT, false},                       // RGBA16UI
     {GL_R11F_G11F_B10F, GL_RGB, GL_UNSIGNED_INT_10F_11F_11F_REV, false},            // R11FG11FB10F
     {GL_RGBA32UI, GL_RGBA_INTEGER, GL_UNSIGNED_INT, false},                         // RGBA32UI
@@ -87,6 +88,7 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format
     {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, false},                             // RG32UI
     {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT, false},                                     // RGBX16F
     {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, false},                             // R32UI
+    {GL_R32I, GL_RED_INTEGER, GL_INT, false},                                       // R32I
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_8X8
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_8X5
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_5X4
@@ -396,6 +398,7 @@ CachedSurfaceView::CachedSurfaceView(CachedSurface& surface, const ViewParams& p
                                      const bool is_proxy)
     : VideoCommon::ViewBase(params), surface{surface}, is_proxy{is_proxy} {
     target = GetTextureTarget(params.target);
+    format = GetFormatTuple(surface.GetSurfaceParams().pixel_format).internal_format;
     if (!is_proxy) {
         texture_view = CreateTextureView();
     }
@@ -405,24 +408,36 @@ CachedSurfaceView::CachedSurfaceView(CachedSurface& surface, const ViewParams& p
 CachedSurfaceView::~CachedSurfaceView() = default;
 
 void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const {
-    ASSERT(params.num_layers == 1 && params.num_levels == 1);
+    ASSERT(params.num_levels == 1);
 
-    const auto& owner_params = surface.GetSurfaceParams();
+    const GLuint texture = surface.GetTexture();
+    if (params.num_layers > 1) {
+        // Layered framebuffer attachments
+        UNIMPLEMENTED_IF(params.base_layer != 0);
 
-    switch (owner_params.target) {
+        switch (params.target) {
+        case SurfaceTarget::Texture2DArray:
+            glFramebufferTexture(target, attachment, texture, params.base_level);
+            break;
+        default:
+            UNIMPLEMENTED();
+        }
+        return;
+    }
+
+    const GLenum view_target = surface.GetTarget();
+    switch (surface.GetSurfaceParams().target) {
     case SurfaceTarget::Texture1D:
-        glFramebufferTexture1D(target, attachment, surface.GetTarget(), surface.GetTexture(),
-                               params.base_level);
+        glFramebufferTexture1D(target, attachment, view_target, texture, params.base_level);
         break;
     case SurfaceTarget::Texture2D:
-        glFramebufferTexture2D(target, attachment, surface.GetTarget(), surface.GetTexture(),
-                               params.base_level);
+        glFramebufferTexture2D(target, attachment, view_target, texture, params.base_level);
         break;
     case SurfaceTarget::Texture1DArray:
     case SurfaceTarget::Texture2DArray:
     case SurfaceTarget::TextureCubemap:
     case SurfaceTarget::TextureCubeArray:
-        glFramebufferTextureLayer(target, attachment, surface.GetTexture(), params.base_level,
+        glFramebufferTextureLayer(target, attachment, texture, params.base_level,
                                   params.base_layer);
         break;
     default:
@@ -454,25 +469,20 @@ void CachedSurfaceView::ApplySwizzle(SwizzleSource x_source, SwizzleSource y_sou
 }
 
 OGLTextureView CachedSurfaceView::CreateTextureView() const {
-    const auto& owner_params = surface.GetSurfaceParams();
     OGLTextureView texture_view;
     texture_view.Create();
 
-    const GLuint handle{texture_view.handle};
-    const FormatTuple& tuple{GetFormatTuple(owner_params.pixel_format)};
-
-    glTextureView(handle, target, surface.texture.handle, tuple.internal_format, params.base_level,
+    glTextureView(texture_view.handle, target, surface.texture.handle, format, params.base_level,
                   params.num_levels, params.base_layer, params.num_layers);
-
-    ApplyTextureDefaults(owner_params, handle);
+    ApplyTextureDefaults(surface.GetSurfaceParams(), texture_view.handle);
 
     return texture_view;
 }
 
 TextureCacheOpenGL::TextureCacheOpenGL(Core::System& system,
                                        VideoCore::RasterizerInterface& rasterizer,
-                                       const Device& device)
-    : TextureCacheBase{system, rasterizer} {
+                                       const Device& device, StateTracker& state_tracker)
+    : TextureCacheBase{system, rasterizer}, state_tracker{state_tracker} {
     src_framebuffer.Create();
     dst_framebuffer.Create();
 }
@@ -506,25 +516,26 @@ void TextureCacheOpenGL::ImageBlit(View& src_view, View& dst_view,
                                    const Tegra::Engines::Fermi2D::Config& copy_config) {
     const auto& src_params{src_view->GetSurfaceParams()};
     const auto& dst_params{dst_view->GetSurfaceParams()};
+    UNIMPLEMENTED_IF(src_params.target == SurfaceTarget::Texture3D);
+    UNIMPLEMENTED_IF(dst_params.target == SurfaceTarget::Texture3D);
 
-    OpenGLState prev_state{OpenGLState::GetCurState()};
-    SCOPE_EXIT({
-        prev_state.AllDirty();
-        prev_state.Apply();
-    });
-
-    OpenGLState state;
-    state.draw.read_framebuffer = src_framebuffer.handle;
-    state.draw.draw_framebuffer = dst_framebuffer.handle;
-    state.framebuffer_srgb.enabled = dst_params.srgb_conversion;
-    state.AllDirty();
-    state.Apply();
+    state_tracker.NotifyScissor0();
+    state_tracker.NotifyFramebuffer();
+    state_tracker.NotifyRasterizeEnable();
+    state_tracker.NotifyFramebufferSRGB();
 
-    u32 buffers{};
+    if (dst_params.srgb_conversion) {
+        glEnable(GL_FRAMEBUFFER_SRGB);
+    } else {
+        glDisable(GL_FRAMEBUFFER_SRGB);
+    }
+    glDisable(GL_RASTERIZER_DISCARD);
+    glDisablei(GL_SCISSOR_TEST, 0);
 
-    UNIMPLEMENTED_IF(src_params.target == SurfaceTarget::Texture3D);
-    UNIMPLEMENTED_IF(dst_params.target == SurfaceTarget::Texture3D);
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, src_framebuffer.handle);
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, dst_framebuffer.handle);
 
+    GLenum buffers = 0;
     if (src_params.type == SurfaceType::ColorTexture) {
         src_view->Attach(GL_COLOR_ATTACHMENT0, GL_READ_FRAMEBUFFER);
         glFramebufferTexture2D(GL_READ_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.h b/src/video_core/renderer_opengl/gl_texture_cache.h
index 8e13ab38b..6658c6ffd 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.h
+++ b/src/video_core/renderer_opengl/gl_texture_cache.h
@@ -27,6 +27,7 @@ using VideoCommon::ViewParams;
 class CachedSurfaceView;
 class CachedSurface;
 class TextureCacheOpenGL;
+class StateTracker;
 
 using Surface = std::shared_ptr<CachedSurface>;
 using View = std::shared_ptr<CachedSurfaceView>;
@@ -96,6 +97,10 @@ public:
         return texture_view.handle;
     }
 
+    GLenum GetFormat() const {
+        return format;
+    }
+
     const SurfaceParams& GetSurfaceParams() const {
         return surface.GetSurfaceParams();
     }
@@ -113,6 +118,7 @@ private:
 
     CachedSurface& surface;
     GLenum target{};
+    GLenum format{};
 
     OGLTextureView texture_view;
     u32 swizzle{};
@@ -122,7 +128,7 @@ private:
 class TextureCacheOpenGL final : public TextureCacheBase {
 public:
     explicit TextureCacheOpenGL(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
-                                const Device& device);
+                                const Device& device, StateTracker& state_tracker);
     ~TextureCacheOpenGL();
 
 protected:
@@ -139,6 +145,8 @@ protected:
 private:
     GLuint FetchPBO(std::size_t buffer_size);
 
+    StateTracker& state_tracker;
+
     OGLFramebuffer src_framebuffer;
     OGLFramebuffer dst_framebuffer;
     std::unordered_map<u32, OGLBuffer> copy_pbo_cache;
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index 7ed505628..89f0e04ef 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -92,8 +92,32 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
         }
     case Maxwell::VertexAttribute::Type::UnsignedScaled:
         switch (attrib.size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
         case Maxwell::VertexAttribute::Size::Size_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
             return GL_UNSIGNED_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return GL_UNSIGNED_SHORT;
+        default:
+            LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
+            return {};
+        }
+    case Maxwell::VertexAttribute::Type::SignedScaled:
+        switch (attrib.size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return GL_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return GL_SHORT;
         default:
             LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
             return {};
@@ -401,24 +425,24 @@ inline GLenum StencilOp(Maxwell::StencilOp stencil) {
     return GL_KEEP;
 }
 
-inline GLenum FrontFace(Maxwell::Cull::FrontFace front_face) {
+inline GLenum FrontFace(Maxwell::FrontFace front_face) {
     switch (front_face) {
-    case Maxwell::Cull::FrontFace::ClockWise:
+    case Maxwell::FrontFace::ClockWise:
         return GL_CW;
-    case Maxwell::Cull::FrontFace::CounterClockWise:
+    case Maxwell::FrontFace::CounterClockWise:
         return GL_CCW;
     }
     LOG_ERROR(Render_OpenGL, "Unimplemented front face cull={}", static_cast<u32>(front_face));
     return GL_CCW;
 }
 
-inline GLenum CullFace(Maxwell::Cull::CullFace cull_face) {
+inline GLenum CullFace(Maxwell::CullFace cull_face) {
     switch (cull_face) {
-    case Maxwell::Cull::CullFace::Front:
+    case Maxwell::CullFace::Front:
         return GL_FRONT;
-    case Maxwell::Cull::CullFace::Back:
+    case Maxwell::CullFace::Back:
         return GL_BACK;
-    case Maxwell::Cull::CullFace::FrontAndBack:
+    case Maxwell::CullFace::FrontAndBack:
         return GL_FRONT_AND_BACK;
     }
     LOG_ERROR(Render_OpenGL, "Unimplemented cull face={}", static_cast<u32>(cull_face));
@@ -464,5 +488,18 @@ inline GLenum LogicOp(Maxwell::LogicOperation operation) {
     return GL_COPY;
 }
 
+inline GLenum PolygonMode(Maxwell::PolygonMode polygon_mode) {
+    switch (polygon_mode) {
+    case Maxwell::PolygonMode::Point:
+        return GL_POINT;
+    case Maxwell::PolygonMode::Line:
+        return GL_LINE;
+    case Maxwell::PolygonMode::Fill:
+        return GL_FILL;
+    }
+    UNREACHABLE_MSG("Invalid polygon mode={}", static_cast<int>(polygon_mode));
+    return GL_FILL;
+}
+
 } // namespace MaxwellToGL
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index bba16afaf..fca5e3ec0 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -5,30 +5,54 @@
 #include <algorithm>
 #include <cstddef>
 #include <cstdlib>
+#include <cstring>
 #include <memory>
+
 #include <glad/glad.h>
+
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/microprofile.h"
 #include "common/telemetry.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_window_context.h"
 #include "core/memory.h"
 #include "core/perf_stats.h"
 #include "core/settings.h"
 #include "core/telemetry_session.h"
 #include "video_core/morton.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
+#include "video_core/renderer_opengl/gl_shader_manager.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
 
 namespace OpenGL {
 
 namespace {
 
-constexpr char vertex_shader[] = R"(
+// If the size of this is too small, it ends up creating a soft cap on FPS as the renderer will have
+// to wait on available presentation frames.
+constexpr std::size_t SWAP_CHAIN_SIZE = 3;
+
+struct Frame {
+    u32 width{};                      /// Width of the frame (to detect resize)
+    u32 height{};                     /// Height of the frame
+    bool color_reloaded{};            /// Texture attachment was recreated (ie: resized)
+    OpenGL::OGLRenderbuffer color{};  /// Buffer shared between the render/present FBO
+    OpenGL::OGLFramebuffer render{};  /// FBO created on the render thread
+    OpenGL::OGLFramebuffer present{}; /// FBO created on the present thread
+    GLsync render_fence{};            /// Fence created on the render thread
+    GLsync present_fence{};           /// Fence created on the presentation thread
+    bool is_srgb{};                   /// Framebuffer is sRGB or RGB
+};
+
+constexpr char VERTEX_SHADER[] = R"(
 #version 430 core
 
+out gl_PerVertex {
+    vec4 gl_Position;
+};
+
 layout (location = 0) in vec2 vert_position;
 layout (location = 1) in vec2 vert_tex_coord;
 layout (location = 0) out vec2 frag_tex_coord;
@@ -49,7 +73,7 @@ void main() {
 }
 )";
 
-constexpr char fragment_shader[] = R"(
+constexpr char FRAGMENT_SHADER[] = R"(
 #version 430 core
 
 layout (location = 0) in vec2 frag_tex_coord;
@@ -58,7 +82,7 @@ layout (location = 0) out vec4 color;
 layout (binding = 0) uniform sampler2D color_texture;
 
 void main() {
-    color = texture(color_texture, frag_tex_coord);
+    color = vec4(texture(color_texture, frag_tex_coord).rgb, 1.0f);
 }
 )";
 
@@ -67,13 +91,31 @@ constexpr GLint TexCoordLocation = 1;
 constexpr GLint ModelViewMatrixLocation = 0;
 
 struct ScreenRectVertex {
-    constexpr ScreenRectVertex(GLfloat x, GLfloat y, GLfloat u, GLfloat v)
-        : position{{x, y}}, tex_coord{{u, v}} {}
+    constexpr ScreenRectVertex(u32 x, u32 y, GLfloat u, GLfloat v)
+        : position{{static_cast<GLfloat>(x), static_cast<GLfloat>(y)}}, tex_coord{{u, v}} {}
 
     std::array<GLfloat, 2> position;
     std::array<GLfloat, 2> tex_coord;
 };
 
+/// Returns true if any debug tool is attached
+bool HasDebugTool() {
+    const bool nsight = std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
+    if (nsight) {
+        return true;
+    }
+
+    GLint num_extensions;
+    glGetIntegerv(GL_NUM_EXTENSIONS, &num_extensions);
+    for (GLuint index = 0; index < static_cast<GLuint>(num_extensions); ++index) {
+        const auto name = reinterpret_cast<const char*>(glGetStringi(GL_EXTENSIONS, index));
+        if (!std::strcmp(name, "GL_EXT_debug_tool")) {
+            return true;
+        }
+    }
+    return false;
+}
+
 /**
  * Defines a 1:1 pixel ortographic projection matrix with (0,0) on the top-left
  * corner and (width, height) on the lower-bottom.
@@ -157,22 +199,229 @@ void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severit
 
 } // Anonymous namespace
 
+/**
+ * For smooth Vsync rendering, we want to always present the latest frame that the core generates,
+ * but also make sure that rendering happens at the pace that the frontend dictates. This is a
+ * helper class that the renderer uses to sync frames between the render thread and the presentation
+ * thread
+ */
+class FrameMailbox {
+public:
+    std::mutex swap_chain_lock;
+    std::condition_variable present_cv;
+    std::array<Frame, SWAP_CHAIN_SIZE> swap_chain{};
+    std::queue<Frame*> free_queue;
+    std::deque<Frame*> present_queue;
+    Frame* previous_frame{};
+
+    FrameMailbox() : has_debug_tool{HasDebugTool()} {
+        for (auto& frame : swap_chain) {
+            free_queue.push(&frame);
+        }
+    }
+
+    ~FrameMailbox() {
+        // lock the mutex and clear out the present and free_queues and notify any people who are
+        // blocked to prevent deadlock on shutdown
+        std::scoped_lock lock{swap_chain_lock};
+        std::queue<Frame*>().swap(free_queue);
+        present_queue.clear();
+        present_cv.notify_all();
+    }
+
+    void ReloadPresentFrame(Frame* frame, u32 height, u32 width) {
+        frame->present.Release();
+        frame->present.Create();
+        GLint previous_draw_fbo{};
+        glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &previous_draw_fbo);
+        glBindFramebuffer(GL_FRAMEBUFFER, frame->present.handle);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate present FBO!");
+        }
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, previous_draw_fbo);
+        frame->color_reloaded = false;
+    }
+
+    void ReloadRenderFrame(Frame* frame, u32 width, u32 height) {
+        // Recreate the color texture attachment
+        frame->color.Release();
+        frame->color.Create();
+        const GLenum internal_format = frame->is_srgb ? GL_SRGB8 : GL_RGB8;
+        glNamedRenderbufferStorage(frame->color.handle, internal_format, width, height);
+
+        // Recreate the FBO for the render target
+        frame->render.Release();
+        frame->render.Create();
+        glBindFramebuffer(GL_FRAMEBUFFER, frame->render.handle);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate render FBO!");
+        }
+
+        frame->width = width;
+        frame->height = height;
+        frame->color_reloaded = true;
+    }
+
+    Frame* GetRenderFrame() {
+        std::unique_lock lock{swap_chain_lock};
+
+        // If theres no free frames, we will reuse the oldest render frame
+        if (free_queue.empty()) {
+            auto frame = present_queue.back();
+            present_queue.pop_back();
+            return frame;
+        }
+
+        Frame* frame = free_queue.front();
+        free_queue.pop();
+        return frame;
+    }
+
+    void ReleaseRenderFrame(Frame* frame) {
+        std::unique_lock lock{swap_chain_lock};
+        present_queue.push_front(frame);
+        present_cv.notify_one();
+
+        DebugNotifyNextFrame();
+    }
+
+    Frame* TryGetPresentFrame(int timeout_ms) {
+        DebugWaitForNextFrame();
+
+        std::unique_lock lock{swap_chain_lock};
+        // wait for new entries in the present_queue
+        present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
+                            [&] { return !present_queue.empty(); });
+        if (present_queue.empty()) {
+            // timed out waiting for a frame to draw so return the previous frame
+            return previous_frame;
+        }
+
+        // free the previous frame and add it back to the free queue
+        if (previous_frame) {
+            free_queue.push(previous_frame);
+        }
+
+        // the newest entries are pushed to the front of the queue
+        Frame* frame = present_queue.front();
+        present_queue.pop_front();
+        // remove all old entries from the present queue and move them back to the free_queue
+        for (auto f : present_queue) {
+            free_queue.push(f);
+        }
+        present_queue.clear();
+        previous_frame = frame;
+        return frame;
+    }
+
+private:
+    std::mutex debug_synch_mutex;
+    std::condition_variable debug_synch_condition;
+    std::atomic_int frame_for_debug{};
+    const bool has_debug_tool; // When true, using a GPU debugger, so keep frames in lock-step
+
+    /// Signal that a new frame is available (called from GPU thread)
+    void DebugNotifyNextFrame() {
+        if (!has_debug_tool) {
+            return;
+        }
+        frame_for_debug++;
+        std::lock_guard lock{debug_synch_mutex};
+        debug_synch_condition.notify_one();
+    }
+
+    /// Wait for a new frame to be available (called from presentation thread)
+    void DebugWaitForNextFrame() {
+        if (!has_debug_tool) {
+            return;
+        }
+        const int last_frame = frame_for_debug;
+        std::unique_lock lock{debug_synch_mutex};
+        debug_synch_condition.wait(lock,
+                                   [this, last_frame] { return frame_for_debug > last_frame; });
+    }
+};
+
 RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system)
-    : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system} {}
+    : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system},
+      frame_mailbox{std::make_unique<FrameMailbox>()} {}
 
 RendererOpenGL::~RendererOpenGL() = default;
 
+MICROPROFILE_DEFINE(OpenGL_RenderFrame, "OpenGL", "Render Frame", MP_RGB(128, 128, 64));
+MICROPROFILE_DEFINE(OpenGL_WaitPresent, "OpenGL", "Wait For Present", MP_RGB(128, 128, 128));
+
 void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
-    // Maintain the rasterizer's state as a priority
-    OpenGLState prev_state = OpenGLState::GetCurState();
-    state.AllDirty();
-    state.Apply();
+    render_window.PollEvents();
+
+    if (!framebuffer) {
+        return;
+    }
+
+    PrepareRendertarget(framebuffer);
+    RenderScreenshot();
+
+    Frame* frame;
+    {
+        MICROPROFILE_SCOPE(OpenGL_WaitPresent);
+
+        frame = frame_mailbox->GetRenderFrame();
+
+        // Clean up sync objects before drawing
+
+        // INTEL driver workaround. We can't delete the previous render sync object until we are
+        // sure that the presentation is done
+        if (frame->present_fence) {
+            glClientWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
+        }
+
+        // delete the draw fence if the frame wasn't presented
+        if (frame->render_fence) {
+            glDeleteSync(frame->render_fence);
+            frame->render_fence = 0;
+        }
+
+        // wait for the presentation to be done
+        if (frame->present_fence) {
+            glWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
+            glDeleteSync(frame->present_fence);
+            frame->present_fence = 0;
+        }
+    }
 
+    {
+        MICROPROFILE_SCOPE(OpenGL_RenderFrame);
+        const auto& layout = render_window.GetFramebufferLayout();
+
+        // Recreate the frame if the size of the window has changed
+        if (layout.width != frame->width || layout.height != frame->height ||
+            screen_info.display_srgb != frame->is_srgb) {
+            LOG_DEBUG(Render_OpenGL, "Reloading render frame");
+            frame->is_srgb = screen_info.display_srgb;
+            frame_mailbox->ReloadRenderFrame(frame, layout.width, layout.height);
+        }
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, frame->render.handle);
+        DrawScreen(layout);
+        // Create a fence for the frontend to wait on and swap this frame to OffTex
+        frame->render_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+        glFlush();
+        frame_mailbox->ReleaseRenderFrame(frame);
+        m_current_frame++;
+        rasterizer->TickFrame();
+    }
+}
+
+void RendererOpenGL::PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer) {
     if (framebuffer) {
         // If framebuffer is provided, reload it from memory to a texture
         if (screen_info.texture.width != static_cast<GLsizei>(framebuffer->width) ||
             screen_info.texture.height != static_cast<GLsizei>(framebuffer->height) ||
-            screen_info.texture.pixel_format != framebuffer->pixel_format) {
+            screen_info.texture.pixel_format != framebuffer->pixel_format ||
+            gl_framebuffer_data.empty()) {
             // Reallocate texture if the framebuffer size has changed.
             // This is expected to not happen very often and hence should not be a
             // performance problem.
@@ -181,22 +430,7 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
 
         // Load the framebuffer from memory, draw it to the screen, and swap buffers
         LoadFBToScreenInfo(*framebuffer);
-
-        if (renderer_settings.screenshot_requested)
-            CaptureScreenshot();
-
-        DrawScreen(render_window.GetFramebufferLayout());
-
-        rasterizer->TickFrame();
-
-        render_window.SwapBuffers();
     }
-
-    render_window.PollEvents();
-
-    // Restore the rasterizer state
-    prev_state.AllDirty();
-    prev_state.Apply();
 }
 
 void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) {
@@ -249,31 +483,24 @@ void RendererOpenGL::InitOpenGLObjects() {
     glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
                  0.0f);
 
-    // Link shaders and get variable locations
-    shader.CreateFromSource(vertex_shader, nullptr, fragment_shader);
-    state.draw.shader_program = shader.handle;
-    state.AllDirty();
-    state.Apply();
+    // Create shader programs
+    OGLShader vertex_shader;
+    vertex_shader.Create(VERTEX_SHADER, GL_VERTEX_SHADER);
+
+    OGLShader fragment_shader;
+    fragment_shader.Create(FRAGMENT_SHADER, GL_FRAGMENT_SHADER);
+
+    vertex_program.Create(true, false, vertex_shader.handle);
+    fragment_program.Create(true, false, fragment_shader.handle);
+
+    // Create program pipeline
+    program_manager.Create();
 
     // Generate VBO handle for drawing
     vertex_buffer.Create();
 
-    // Generate VAO
-    vertex_array.Create();
-    state.draw.vertex_array = vertex_array.handle;
-
     // Attach vertex data to VAO
     glNamedBufferData(vertex_buffer.handle, sizeof(ScreenRectVertex) * 4, nullptr, GL_STREAM_DRAW);
-    glVertexArrayAttribFormat(vertex_array.handle, PositionLocation, 2, GL_FLOAT, GL_FALSE,
-                              offsetof(ScreenRectVertex, position));
-    glVertexArrayAttribFormat(vertex_array.handle, TexCoordLocation, 2, GL_FLOAT, GL_FALSE,
-                              offsetof(ScreenRectVertex, tex_coord));
-    glVertexArrayAttribBinding(vertex_array.handle, PositionLocation, 0);
-    glVertexArrayAttribBinding(vertex_array.handle, TexCoordLocation, 0);
-    glEnableVertexArrayAttrib(vertex_array.handle, PositionLocation);
-    glEnableVertexArrayAttrib(vertex_array.handle, TexCoordLocation);
-    glVertexArrayVertexBuffer(vertex_array.handle, 0, vertex_buffer.handle, 0,
-                              sizeof(ScreenRectVertex));
 
     // Allocate textures for the screen
     screen_info.texture.resource.Create(GL_TEXTURE_2D);
@@ -306,7 +533,8 @@ void RendererOpenGL::CreateRasterizer() {
     if (rasterizer) {
         return;
     }
-    rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, screen_info);
+    rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, screen_info,
+                                                    program_manager, state_tracker);
 }
 
 void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
@@ -345,8 +573,19 @@ void RendererOpenGL::ConfigureFramebufferTexture(TextureInfo& texture,
     glTextureStorage2D(texture.resource.handle, 1, internal_format, texture.width, texture.height);
 }
 
-void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x, float y, float w,
-                                         float h) {
+void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
+    if (renderer_settings.set_background_color) {
+        // Update background color before drawing
+        glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
+                     0.0f);
+    }
+
+    // Set projection matrix
+    const std::array ortho_matrix =
+        MakeOrthographicMatrix(static_cast<float>(layout.width), static_cast<float>(layout.height));
+    glProgramUniformMatrix3x2fv(vertex_program.handle, ModelViewMatrixLocation, 1, GL_FALSE,
+                                std::data(ortho_matrix));
+
     const auto& texcoords = screen_info.display_texcoords;
     auto left = texcoords.left;
     auto right = texcoords.right;
@@ -378,60 +617,129 @@ void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x,
                   static_cast<f32>(screen_info.texture.height);
     }
 
+    const auto& screen = layout.screen;
     const std::array vertices = {
-        ScreenRectVertex(x, y, texcoords.top * scale_u, left * scale_v),
-        ScreenRectVertex(x + w, y, texcoords.bottom * scale_u, left * scale_v),
-        ScreenRectVertex(x, y + h, texcoords.top * scale_u, right * scale_v),
-        ScreenRectVertex(x + w, y + h, texcoords.bottom * scale_u, right * scale_v),
+        ScreenRectVertex(screen.left, screen.top, texcoords.top * scale_u, left * scale_v),
+        ScreenRectVertex(screen.right, screen.top, texcoords.bottom * scale_u, left * scale_v),
+        ScreenRectVertex(screen.left, screen.bottom, texcoords.top * scale_u, right * scale_v),
+        ScreenRectVertex(screen.right, screen.bottom, texcoords.bottom * scale_u, right * scale_v),
     };
-
-    state.textures[0] = screen_info.display_texture;
-    state.framebuffer_srgb.enabled = screen_info.display_srgb;
-    state.AllDirty();
-    state.Apply();
     glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), std::data(vertices));
-    glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
-    // Restore default state
-    state.framebuffer_srgb.enabled = false;
-    state.textures[0] = 0;
-    state.AllDirty();
-    state.Apply();
-}
 
-void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
-    if (renderer_settings.set_background_color) {
-        // Update background color before drawing
-        glClearColor(Settings::values.bg_red, Settings::values.bg_green, Settings::values.bg_blue,
-                     0.0f);
+    // TODO: Signal state tracker about these changes
+    state_tracker.NotifyScreenDrawVertexArray();
+    state_tracker.NotifyPolygonModes();
+    state_tracker.NotifyViewport0();
+    state_tracker.NotifyScissor0();
+    state_tracker.NotifyColorMask0();
+    state_tracker.NotifyBlend0();
+    state_tracker.NotifyFramebuffer();
+    state_tracker.NotifyFrontFace();
+    state_tracker.NotifyCullTest();
+    state_tracker.NotifyDepthTest();
+    state_tracker.NotifyStencilTest();
+    state_tracker.NotifyPolygonOffset();
+    state_tracker.NotifyRasterizeEnable();
+    state_tracker.NotifyFramebufferSRGB();
+    state_tracker.NotifyLogicOp();
+    state_tracker.NotifyClipControl();
+    state_tracker.NotifyAlphaTest();
+
+    program_manager.UseVertexShader(vertex_program.handle);
+    program_manager.UseGeometryShader(0);
+    program_manager.UseFragmentShader(fragment_program.handle);
+    program_manager.BindGraphicsPipeline();
+
+    glEnable(GL_CULL_FACE);
+    if (screen_info.display_srgb) {
+        glEnable(GL_FRAMEBUFFER_SRGB);
+    } else {
+        glDisable(GL_FRAMEBUFFER_SRGB);
     }
+    glDisable(GL_COLOR_LOGIC_OP);
+    glDisable(GL_DEPTH_TEST);
+    glDisable(GL_STENCIL_TEST);
+    glDisable(GL_POLYGON_OFFSET_FILL);
+    glDisable(GL_RASTERIZER_DISCARD);
+    glDisable(GL_ALPHA_TEST);
+    glDisablei(GL_BLEND, 0);
+    glDisablei(GL_SCISSOR_TEST, 0);
+    glPolygonMode(GL_FRONT_AND_BACK, GL_FILL);
+    glCullFace(GL_BACK);
+    glFrontFace(GL_CW);
+    glColorMaski(0, GL_TRUE, GL_TRUE, GL_TRUE, GL_TRUE);
+    glClipControl(GL_LOWER_LEFT, GL_ZERO_TO_ONE);
+    glViewportIndexedf(0, 0.0f, 0.0f, static_cast<GLfloat>(layout.width),
+                       static_cast<GLfloat>(layout.height));
+    glDepthRangeIndexed(0, 0.0, 0.0);
+
+    glEnableVertexAttribArray(PositionLocation);
+    glEnableVertexAttribArray(TexCoordLocation);
+    glVertexAttribDivisor(PositionLocation, 0);
+    glVertexAttribDivisor(TexCoordLocation, 0);
+    glVertexAttribFormat(PositionLocation, 2, GL_FLOAT, GL_FALSE,
+                         offsetof(ScreenRectVertex, position));
+    glVertexAttribFormat(TexCoordLocation, 2, GL_FLOAT, GL_FALSE,
+                         offsetof(ScreenRectVertex, tex_coord));
+    glVertexAttribBinding(PositionLocation, 0);
+    glVertexAttribBinding(TexCoordLocation, 0);
+    glBindVertexBuffer(0, vertex_buffer.handle, 0, sizeof(ScreenRectVertex));
+
+    glBindTextureUnit(0, screen_info.display_texture);
+    glBindSampler(0, 0);
 
-    const auto& screen = layout.screen;
-
-    glViewport(0, 0, layout.width, layout.height);
     glClear(GL_COLOR_BUFFER_BIT);
+    glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
+}
 
-    // Set projection matrix
-    const std::array ortho_matrix =
-        MakeOrthographicMatrix(static_cast<float>(layout.width), static_cast<float>(layout.height));
-    glUniformMatrix3x2fv(ModelViewMatrixLocation, 1, GL_FALSE, ortho_matrix.data());
+void RendererOpenGL::TryPresent(int timeout_ms) {
+    const auto& layout = render_window.GetFramebufferLayout();
+    auto frame = frame_mailbox->TryGetPresentFrame(timeout_ms);
+    if (!frame) {
+        LOG_DEBUG(Render_OpenGL, "TryGetPresentFrame returned no frame to present");
+        return;
+    }
 
-    DrawScreenTriangles(screen_info, static_cast<float>(screen.left),
-                        static_cast<float>(screen.top), static_cast<float>(screen.GetWidth()),
-                        static_cast<float>(screen.GetHeight()));
+    // Clearing before a full overwrite of a fbo can signal to drivers that they can avoid a
+    // readback since we won't be doing any blending
+    glClear(GL_COLOR_BUFFER_BIT);
 
-    m_current_frame++;
+    // Recreate the presentation FBO if the color attachment was changed
+    if (frame->color_reloaded) {
+        LOG_DEBUG(Render_OpenGL, "Reloading present frame");
+        frame_mailbox->ReloadPresentFrame(frame, layout.width, layout.height);
+    }
+    glWaitSync(frame->render_fence, 0, GL_TIMEOUT_IGNORED);
+    // INTEL workaround.
+    // Normally we could just delete the draw fence here, but due to driver bugs, we can just delete
+    // it on the emulation thread without too much penalty
+    // glDeleteSync(frame.render_sync);
+    // frame.render_sync = 0;
+
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, frame->present.handle);
+    glBlitFramebuffer(0, 0, frame->width, frame->height, 0, 0, layout.width, layout.height,
+                      GL_COLOR_BUFFER_BIT, GL_LINEAR);
+
+    // Insert fence for the main thread to block on
+    frame->present_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+    glFlush();
+
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
 }
 
-void RendererOpenGL::UpdateFramerate() {}
+void RendererOpenGL::RenderScreenshot() {
+    if (!renderer_settings.screenshot_requested) {
+        return;
+    }
+
+    GLint old_read_fb;
+    GLint old_draw_fb;
+    glGetIntegerv(GL_READ_FRAMEBUFFER_BINDING, &old_read_fb);
+    glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &old_draw_fb);
 
-void RendererOpenGL::CaptureScreenshot() {
     // Draw the current frame to the screenshot framebuffer
     screenshot_framebuffer.Create();
-    GLuint old_read_fb = state.draw.read_framebuffer;
-    GLuint old_draw_fb = state.draw.draw_framebuffer;
-    state.draw.read_framebuffer = state.draw.draw_framebuffer = screenshot_framebuffer.handle;
-    state.AllDirty();
-    state.Apply();
+    glBindFramebuffer(GL_FRAMEBUFFER, screenshot_framebuffer.handle);
 
     Layout::FramebufferLayout layout{renderer_settings.screenshot_framebuffer_layout};
 
@@ -448,19 +756,16 @@ void RendererOpenGL::CaptureScreenshot() {
                  renderer_settings.screenshot_bits);
 
     screenshot_framebuffer.Release();
-    state.draw.read_framebuffer = old_read_fb;
-    state.draw.draw_framebuffer = old_draw_fb;
-    state.AllDirty();
-    state.Apply();
     glDeleteRenderbuffers(1, &renderbuffer);
 
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, old_read_fb);
+    glBindFramebuffer(GL_DRAW_FRAMEBUFFER, old_draw_fb);
+
     renderer_settings.screenshot_complete_callback();
     renderer_settings.screenshot_requested = false;
 }
 
 bool RendererOpenGL::Init() {
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{render_window};
-
     if (GLAD_GL_KHR_debug) {
         glEnable(GL_DEBUG_OUTPUT);
         glDebugMessageCallback(DebugHandler, nullptr);
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h
index b56328a7f..33073ce5b 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.h
+++ b/src/video_core/renderer_opengl/renderer_opengl.h
@@ -10,7 +10,8 @@
 #include "common/math_util.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
-#include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/renderer_opengl/gl_shader_manager.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 
 namespace Core {
 class System;
@@ -44,19 +45,23 @@ struct ScreenInfo {
     TextureInfo texture;
 };
 
+struct PresentationTexture {
+    u32 width = 0;
+    u32 height = 0;
+    OGLTexture texture;
+};
+
+class FrameMailbox;
+
 class RendererOpenGL final : public VideoCore::RendererBase {
 public:
     explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system);
     ~RendererOpenGL() override;
 
-    /// Swap buffers (render frame)
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-
-    /// Initialize the renderer
     bool Init() override;
-
-    /// Shutdown the renderer
     void ShutDown() override;
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
+    void TryPresent(int timeout_ms) override;
 
 private:
     /// Initializes the OpenGL state and creates persistent objects.
@@ -72,12 +77,7 @@ private:
     /// Draws the emulated screens to the emulator window.
     void DrawScreen(const Layout::FramebufferLayout& layout);
 
-    void DrawScreenTriangles(const ScreenInfo& screen_info, float x, float y, float w, float h);
-
-    /// Updates the framerate.
-    void UpdateFramerate();
-
-    void CaptureScreenshot();
+    void RenderScreenshot();
 
     /// Loads framebuffer from emulated memory into the active OpenGL texture.
     void LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer);
@@ -87,26 +87,34 @@ private:
     void LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, u8 color_a,
                                     const TextureInfo& texture);
 
+    void PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer);
+
     Core::Frontend::EmuWindow& emu_window;
     Core::System& system;
 
-    OpenGLState state;
+    StateTracker state_tracker{system};
 
     // OpenGL object IDs
-    OGLVertexArray vertex_array;
     OGLBuffer vertex_buffer;
-    OGLProgram shader;
+    OGLProgram vertex_program;
+    OGLProgram fragment_program;
     OGLFramebuffer screenshot_framebuffer;
 
     /// Display information for Switch screen
     ScreenInfo screen_info;
 
+    /// Global dummy shader pipeline
+    GLShader::ProgramManager program_manager;
+
     /// OpenGL framebuffer data
     std::vector<u8> gl_framebuffer_data;
 
     /// Used for transforming the framebuffer orientation
     Tegra::FramebufferConfig::TransformFlags framebuffer_transform_flags;
     Common::Rectangle<int> framebuffer_crop_rect;
+
+    /// Frame presentation mailbox
+    std::unique_ptr<FrameMailbox> frame_mailbox;
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/utils.cpp b/src/video_core/renderer_opengl/utils.cpp
index ac99e6385..b751086fa 100644
--- a/src/video_core/renderer_opengl/utils.cpp
+++ b/src/video_core/renderer_opengl/utils.cpp
@@ -9,6 +9,7 @@
 #include <glad/glad.h>
 
 #include "common/common_types.h"
+#include "video_core/renderer_opengl/gl_state_tracker.h"
 #include "video_core/renderer_opengl/utils.h"
 
 namespace OpenGL {
@@ -20,12 +21,12 @@ struct VertexArrayPushBuffer::Entry {
     GLsizei stride{};
 };
 
-VertexArrayPushBuffer::VertexArrayPushBuffer() = default;
+VertexArrayPushBuffer::VertexArrayPushBuffer(StateTracker& state_tracker)
+    : state_tracker{state_tracker} {}
 
 VertexArrayPushBuffer::~VertexArrayPushBuffer() = default;
 
-void VertexArrayPushBuffer::Setup(GLuint vao_) {
-    vao = vao_;
+void VertexArrayPushBuffer::Setup() {
     index_buffer = nullptr;
     vertex_buffers.clear();
 }
@@ -41,13 +42,11 @@ void VertexArrayPushBuffer::SetVertexBuffer(GLuint binding_index, const GLuint*
 
 void VertexArrayPushBuffer::Bind() {
     if (index_buffer) {
-        glVertexArrayElementBuffer(vao, *index_buffer);
+        state_tracker.BindIndexBuffer(*index_buffer);
     }
 
-    // TODO(Rodrigo): Find a way to ARB_multi_bind this
     for (const auto& entry : vertex_buffers) {
-        glVertexArrayVertexBuffer(vao, entry.binding_index, *entry.buffer, entry.offset,
-                                  entry.stride);
+        glBindVertexBuffer(entry.binding_index, *entry.buffer, entry.offset, entry.stride);
     }
 }
 
diff --git a/src/video_core/renderer_opengl/utils.h b/src/video_core/renderer_opengl/utils.h
index 3ad7c02d4..47ee3177b 100644
--- a/src/video_core/renderer_opengl/utils.h
+++ b/src/video_core/renderer_opengl/utils.h
@@ -11,12 +11,14 @@
 
 namespace OpenGL {
 
+class StateTracker;
+
 class VertexArrayPushBuffer final {
 public:
-    explicit VertexArrayPushBuffer();
+    explicit VertexArrayPushBuffer(StateTracker& state_tracker);
     ~VertexArrayPushBuffer();
 
-    void Setup(GLuint vao_);
+    void Setup();
 
     void SetIndexBuffer(const GLuint* buffer);
 
@@ -28,7 +30,8 @@ public:
 private:
     struct Entry;
 
-    GLuint vao{};
+    StateTracker& state_tracker;
+
     const GLuint* index_buffer{};
     std::vector<Entry> vertex_buffers;
 };
diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
index 4e3ff231e..2bb376555 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.cpp
@@ -112,19 +112,18 @@ constexpr FixedPipelineState::Rasterizer GetRasterizerState(const Maxwell& regs)
     const auto& clip = regs.view_volume_clip_control;
     const bool depth_clamp_enabled = clip.depth_clamp_near == 1 || clip.depth_clamp_far == 1;
 
-    Maxwell::Cull::FrontFace front_face = regs.cull.front_face;
+    Maxwell::FrontFace front_face = regs.front_face;
     if (regs.screen_y_control.triangle_rast_flip != 0 &&
         regs.viewport_transform[0].scale_y > 0.0f) {
-        if (front_face == Maxwell::Cull::FrontFace::CounterClockWise)
-            front_face = Maxwell::Cull::FrontFace::ClockWise;
-        else if (front_face == Maxwell::Cull::FrontFace::ClockWise)
-            front_face = Maxwell::Cull::FrontFace::CounterClockWise;
+        if (front_face == Maxwell::FrontFace::CounterClockWise)
+            front_face = Maxwell::FrontFace::ClockWise;
+        else if (front_face == Maxwell::FrontFace::ClockWise)
+            front_face = Maxwell::FrontFace::CounterClockWise;
     }
 
     const bool gl_ndc = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
-    return FixedPipelineState::Rasterizer(regs.cull.enabled, depth_bias_enabled,
-                                          depth_clamp_enabled, gl_ndc, regs.cull.cull_face,
-                                          front_face);
+    return FixedPipelineState::Rasterizer(regs.cull_test_enabled, depth_bias_enabled,
+                                          depth_clamp_enabled, gl_ndc, regs.cull_face, front_face);
 }
 
 } // Anonymous namespace
diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.h b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
index 87056ef37..4c8ba7f90 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.h
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
@@ -171,8 +171,8 @@ struct FixedPipelineState {
 
     struct Rasterizer {
         constexpr Rasterizer(bool cull_enable, bool depth_bias_enable, bool depth_clamp_enable,
-                             bool ndc_minus_one_to_one, Maxwell::Cull::CullFace cull_face,
-                             Maxwell::Cull::FrontFace front_face)
+                             bool ndc_minus_one_to_one, Maxwell::CullFace cull_face,
+                             Maxwell::FrontFace front_face)
             : cull_enable{cull_enable}, depth_bias_enable{depth_bias_enable},
               depth_clamp_enable{depth_clamp_enable}, ndc_minus_one_to_one{ndc_minus_one_to_one},
               cull_face{cull_face}, front_face{front_face} {}
@@ -182,8 +182,8 @@ struct FixedPipelineState {
         bool depth_bias_enable;
         bool depth_clamp_enable;
         bool ndc_minus_one_to_one;
-        Maxwell::Cull::CullFace cull_face;
-        Maxwell::Cull::FrontFace front_face;
+        Maxwell::CullFace cull_face;
+        Maxwell::FrontFace front_face;
 
         std::size_t Hash() const noexcept;
 
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 5403c3ab7..f93447610 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -120,11 +120,12 @@ struct FormatTuple {
     {vk::Format::eA8B8G8R8UintPack32, Attachable | Storage},     // ABGR8UI
     {vk::Format::eB5G6R5UnormPack16, {}},                        // B5G6R5U
     {vk::Format::eA2B10G10R10UnormPack32, Attachable | Storage}, // A2B10G10R10U
-    {vk::Format::eA1R5G5B5UnormPack16, Attachable | Storage},    // A1B5G5R5U (flipped with swizzle)
+    {vk::Format::eA1R5G5B5UnormPack16, Attachable},              // A1B5G5R5U (flipped with swizzle)
     {vk::Format::eR8Unorm, Attachable | Storage},                // R8U
     {vk::Format::eR8Uint, Attachable | Storage},                 // R8UI
     {vk::Format::eR16G16B16A16Sfloat, Attachable | Storage},     // RGBA16F
     {vk::Format::eR16G16B16A16Unorm, Attachable | Storage},      // RGBA16U
+    {vk::Format::eR16G16B16A16Snorm, Attachable | Storage},      // RGBA16S
     {vk::Format::eR16G16B16A16Uint, Attachable | Storage},       // RGBA16UI
     {vk::Format::eB10G11R11UfloatPack32, Attachable | Storage},  // R11FG11FB10F
     {vk::Format::eR32G32B32A32Uint, Attachable | Storage},       // RGBA32UI
@@ -159,6 +160,7 @@ struct FormatTuple {
     {vk::Format::eR32G32Uint, Attachable | Storage},             // RG32UI
     {vk::Format::eUndefined, {}},                                // RGBX16F
     {vk::Format::eR32Uint, Attachable | Storage},                // R32UI
+    {vk::Format::eR32Sint, Attachable | Storage},                // R32I
     {vk::Format::eAstc8x8UnormBlock, {}},                        // ASTC_2D_8X8
     {vk::Format::eUndefined, {}},                                // ASTC_2D_8X5
     {vk::Format::eUndefined, {}},                                // ASTC_2D_5X4
@@ -255,6 +257,8 @@ vk::ShaderStageFlagBits ShaderStage(Tegra::Engines::ShaderType stage) {
         return vk::ShaderStageFlagBits::eGeometry;
     case Tegra::Engines::ShaderType::Fragment:
         return vk::ShaderStageFlagBits::eFragment;
+    case Tegra::Engines::ShaderType::Compute:
+        return vk::ShaderStageFlagBits::eCompute;
     }
     UNIMPLEMENTED_MSG("Unimplemented shader stage={}", static_cast<u32>(stage));
     return {};
@@ -330,6 +334,8 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
             return vk::Format::eR16G16B16Unorm;
         case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
             return vk::Format::eR16G16B16A16Unorm;
+        case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
+            return vk::Format::eA2B10G10R10UnormPack32;
         default:
             break;
         }
@@ -363,6 +369,10 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
             return vk::Format::eR8G8B8A8Uint;
         case Maxwell::VertexAttribute::Size::Size_32:
             return vk::Format::eR32Uint;
+        case Maxwell::VertexAttribute::Size::Size_32_32:
+            return vk::Format::eR32G32Uint;
+        case Maxwell::VertexAttribute::Size::Size_32_32_32:
+            return vk::Format::eR32G32B32Uint;
         case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
             return vk::Format::eR32G32B32A32Uint;
         default:
@@ -370,8 +380,22 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
         }
     case Maxwell::VertexAttribute::Type::UnsignedScaled:
         switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
+            return vk::Format::eR8Uscaled;
         case Maxwell::VertexAttribute::Size::Size_8_8:
             return vk::Format::eR8G8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+            return vk::Format::eR8G8B8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return vk::Format::eR8G8B8A8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16:
+            return vk::Format::eR16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+            return vk::Format::eR16G16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+            return vk::Format::eR16G16B16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return vk::Format::eR16G16B16A16Uscaled;
         default:
             break;
         }
@@ -571,24 +595,24 @@ vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor) {
     return {};
 }
 
-vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face) {
+vk::FrontFace FrontFace(Maxwell::FrontFace front_face) {
     switch (front_face) {
-    case Maxwell::Cull::FrontFace::ClockWise:
+    case Maxwell::FrontFace::ClockWise:
         return vk::FrontFace::eClockwise;
-    case Maxwell::Cull::FrontFace::CounterClockWise:
+    case Maxwell::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) {
+vk::CullModeFlags CullFace(Maxwell::CullFace cull_face) {
     switch (cull_face) {
-    case Maxwell::Cull::CullFace::Front:
+    case Maxwell::CullFace::Front:
         return vk::CullModeFlagBits::eFront;
-    case Maxwell::Cull::CullFace::Back:
+    case Maxwell::CullFace::Back:
         return vk::CullModeFlagBits::eBack;
-    case Maxwell::Cull::CullFace::FrontAndBack:
+    case Maxwell::CullFace::FrontAndBack:
         return vk::CullModeFlagBits::eFrontAndBack;
     }
     UNIMPLEMENTED_MSG("Unimplemented cull face={}", static_cast<u32>(cull_face));
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.h b/src/video_core/renderer_vulkan/maxwell_to_vk.h
index 7e9678b7b..24f6ab544 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.h
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.h
@@ -54,9 +54,9 @@ vk::BlendOp BlendEquation(Maxwell::Blend::Equation equation);
 
 vk::BlendFactor BlendFactor(Maxwell::Blend::Factor factor);
 
-vk::FrontFace FrontFace(Maxwell::Cull::FrontFace front_face);
+vk::FrontFace FrontFace(Maxwell::FrontFace front_face);
 
-vk::CullModeFlags CullFace(Maxwell::Cull::CullFace cull_face);
+vk::CullModeFlags CullFace(Maxwell::CullFace cull_face);
 
 vk::ComponentSwizzle SwizzleSource(Tegra::Texture::SwizzleSource swizzle);
 
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
index d5032b432..42bb01418 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
@@ -27,6 +27,7 @@
 #include "video_core/renderer_vulkan/vk_rasterizer.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
 #include "video_core/renderer_vulkan/vk_swapchain.h"
 
 namespace Vulkan {
@@ -106,8 +107,14 @@ RendererVulkan::~RendererVulkan() {
 }
 
 void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
+    render_window.PollEvents();
+
+    if (!framebuffer) {
+        return;
+    }
+
     const auto& layout = render_window.GetFramebufferLayout();
-    if (framebuffer && layout.width > 0 && layout.height > 0 && render_window.IsShown()) {
+    if (layout.width > 0 && layout.height > 0 && render_window.IsShown()) {
         const VAddr framebuffer_addr = framebuffer->address + framebuffer->offset;
         const bool use_accelerated =
             rasterizer->AccelerateDisplay(*framebuffer, framebuffer_addr, framebuffer->stride);
@@ -128,13 +135,16 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
             blit_screen->Recreate();
         }
 
-        render_window.SwapBuffers();
         rasterizer->TickFrame();
     }
 
     render_window.PollEvents();
 }
 
+void RendererVulkan::TryPresent(int /*timeout_ms*/) {
+    // TODO (bunnei): ImplementMe
+}
+
 bool RendererVulkan::Init() {
     PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{};
     render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface);
@@ -168,10 +178,13 @@ bool RendererVulkan::Init() {
     swapchain = std::make_unique<VKSwapchain>(surface, *device);
     swapchain->Create(framebuffer.width, framebuffer.height, false);
 
-    scheduler = std::make_unique<VKScheduler>(*device, *resource_manager);
+    state_tracker = std::make_unique<StateTracker>(system);
+
+    scheduler = std::make_unique<VKScheduler>(*device, *resource_manager, *state_tracker);
 
     rasterizer = std::make_unique<RasterizerVulkan>(system, render_window, screen_info, *device,
-                                                    *resource_manager, *memory_manager, *scheduler);
+                                                    *resource_manager, *memory_manager,
+                                                    *state_tracker, *scheduler);
 
     blit_screen = std::make_unique<VKBlitScreen>(system, render_window, *rasterizer, *device,
                                                  *resource_manager, *memory_manager, *swapchain,
@@ -262,4 +275,4 @@ void RendererVulkan::Report() const {
     telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions);
 }
 
-} // namespace Vulkan
\ No newline at end of file
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h
index a472c5dc9..3da08d2e4 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.h
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.h
@@ -4,8 +4,10 @@
 
 #pragma once
 
+#include <memory>
 #include <optional>
 #include <vector>
+
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_vulkan/declarations.h"
 
@@ -15,6 +17,7 @@ class System;
 
 namespace Vulkan {
 
+class StateTracker;
 class VKBlitScreen;
 class VKDevice;
 class VKFence;
@@ -36,14 +39,10 @@ public:
     explicit RendererVulkan(Core::Frontend::EmuWindow& window, Core::System& system);
     ~RendererVulkan() override;
 
-    /// Swap buffers (render frame)
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-
-    /// Initialize the renderer
     bool Init() override;
-
-    /// Shutdown the renderer
     void ShutDown() override;
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
+    void TryPresent(int timeout_ms) override;
 
 private:
     std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(
@@ -65,6 +64,7 @@ private:
     std::unique_ptr<VKSwapchain> swapchain;
     std::unique_ptr<VKMemoryManager> memory_manager;
     std::unique_ptr<VKResourceManager> resource_manager;
+    std::unique_ptr<StateTracker> state_tracker;
     std::unique_ptr<VKScheduler> scheduler;
     std::unique_ptr<VKBlitScreen> blit_screen;
 };
diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
index 9d5b8de7a..60f57d83e 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
@@ -73,7 +73,7 @@ UniqueDescriptorUpdateTemplate VKComputePipeline::CreateDescriptorUpdateTemplate
     std::vector<vk::DescriptorUpdateTemplateEntry> template_entries;
     u32 binding = 0;
     u32 offset = 0;
-    FillDescriptorUpdateTemplateEntries(device, entries, binding, offset, template_entries);
+    FillDescriptorUpdateTemplateEntries(entries, binding, offset, template_entries);
     if (template_entries.empty()) {
         // If the shader doesn't use descriptor sets, skip template creation.
         return UniqueDescriptorUpdateTemplate{};
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index d1da4f9d3..28d2fbc4f 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -107,8 +107,7 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
     features.occlusionQueryPrecise = true;
     features.fragmentStoresAndAtomics = true;
     features.shaderImageGatherExtended = true;
-    features.shaderStorageImageReadWithoutFormat =
-        is_shader_storage_img_read_without_format_supported;
+    features.shaderStorageImageReadWithoutFormat = is_formatless_image_load_supported;
     features.shaderStorageImageWriteWithoutFormat = true;
     features.textureCompressionASTC_LDR = is_optimal_astc_supported;
 
@@ -148,6 +147,15 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
         LOG_INFO(Render_Vulkan, "Device doesn't support uint8 indexes");
     }
 
+    vk::PhysicalDeviceTransformFeedbackFeaturesEXT transform_feedback;
+    if (ext_transform_feedback) {
+        transform_feedback.transformFeedback = true;
+        transform_feedback.geometryStreams = true;
+        SetNext(next, transform_feedback);
+    } else {
+        LOG_INFO(Render_Vulkan, "Device doesn't support transform feedbacks");
+    }
+
     if (!ext_depth_range_unrestricted) {
         LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted");
     }
@@ -385,7 +393,7 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
         }
     };
 
-    extensions.reserve(14);
+    extensions.reserve(15);
     extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
     extensions.push_back(VK_KHR_16BIT_STORAGE_EXTENSION_NAME);
     extensions.push_back(VK_KHR_8BIT_STORAGE_EXTENSION_NAME);
@@ -397,18 +405,22 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
 
     [[maybe_unused]] const bool nsight =
         std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
-    bool khr_shader_float16_int8{};
-    bool ext_subgroup_size_control{};
+    bool has_khr_shader_float16_int8{};
+    bool has_ext_subgroup_size_control{};
+    bool has_ext_transform_feedback{};
     for (const auto& extension : physical.enumerateDeviceExtensionProperties(nullptr, dldi)) {
         Test(extension, khr_uniform_buffer_standard_layout,
              VK_KHR_UNIFORM_BUFFER_STANDARD_LAYOUT_EXTENSION_NAME, true);
-        Test(extension, khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME, false);
+        Test(extension, has_khr_shader_float16_int8, VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME,
+             false);
         Test(extension, ext_depth_range_unrestricted,
              VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME, true);
         Test(extension, ext_index_type_uint8, VK_EXT_INDEX_TYPE_UINT8_EXTENSION_NAME, true);
         Test(extension, ext_shader_viewport_index_layer,
              VK_EXT_SHADER_VIEWPORT_INDEX_LAYER_EXTENSION_NAME, true);
-        Test(extension, ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME,
+        Test(extension, has_ext_subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME,
+             false);
+        Test(extension, has_ext_transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME,
              false);
         if (Settings::values.renderer_debug) {
             Test(extension, nv_device_diagnostic_checkpoints,
@@ -416,13 +428,13 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
         }
     }
 
-    if (khr_shader_float16_int8) {
+    if (has_khr_shader_float16_int8) {
         is_float16_supported =
             GetFeatures<vk::PhysicalDeviceFloat16Int8FeaturesKHR>(physical, dldi).shaderFloat16;
         extensions.push_back(VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME);
     }
 
-    if (ext_subgroup_size_control) {
+    if (has_ext_subgroup_size_control) {
         const auto features =
             GetFeatures<vk::PhysicalDeviceSubgroupSizeControlFeaturesEXT>(physical, dldi);
         const auto properties =
@@ -439,6 +451,20 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
         is_warp_potentially_bigger = true;
     }
 
+    if (has_ext_transform_feedback) {
+        const auto features =
+            GetFeatures<vk::PhysicalDeviceTransformFeedbackFeaturesEXT>(physical, dldi);
+        const auto properties =
+            GetProperties<vk::PhysicalDeviceTransformFeedbackPropertiesEXT>(physical, dldi);
+
+        if (features.transformFeedback && features.geometryStreams &&
+            properties.maxTransformFeedbackStreams >= 4 && properties.maxTransformFeedbackBuffers &&
+            properties.transformFeedbackQueries && properties.transformFeedbackDraw) {
+            extensions.push_back(VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME);
+            ext_transform_feedback = true;
+        }
+    }
+
     return extensions;
 }
 
@@ -467,8 +493,7 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
 
 void VKDevice::SetupFeatures(const vk::DispatchLoaderDynamic& dldi) {
     const auto supported_features{physical.getFeatures(dldi)};
-    is_shader_storage_img_read_without_format_supported =
-        supported_features.shaderStorageImageReadWithoutFormat;
+    is_formatless_image_load_supported = supported_features.shaderStorageImageReadWithoutFormat;
     is_optimal_astc_supported = IsOptimalAstcSupported(supported_features, dldi);
 }
 
@@ -510,6 +535,7 @@ std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperti
                                         vk::Format::eR32G32Sfloat,
                                         vk::Format::eR32G32Uint,
                                         vk::Format::eR16G16B16A16Uint,
+                                        vk::Format::eR16G16B16A16Snorm,
                                         vk::Format::eR16G16B16A16Unorm,
                                         vk::Format::eR16G16Unorm,
                                         vk::Format::eR16G16Snorm,
@@ -523,6 +549,7 @@ std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperti
                                         vk::Format::eB10G11R11UfloatPack32,
                                         vk::Format::eR32Sfloat,
                                         vk::Format::eR32Uint,
+                                        vk::Format::eR32Sint,
                                         vk::Format::eR16Sfloat,
                                         vk::Format::eR16G16B16A16Sfloat,
                                         vk::Format::eB8G8R8A8Unorm,
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index 2c27ad730..6e656517f 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -122,11 +122,6 @@ public:
         return properties.limits.maxPushConstantsSize;
     }
 
-    /// Returns true if Shader storage Image Read Without Format supported.
-    bool IsShaderStorageImageReadWithoutFormatSupported() const {
-        return is_shader_storage_img_read_without_format_supported;
-    }
-
     /// Returns true if ASTC is natively supported.
     bool IsOptimalAstcSupported() const {
         return is_optimal_astc_supported;
@@ -147,6 +142,11 @@ public:
         return (guest_warp_stages & stage) != vk::ShaderStageFlags{};
     }
 
+    /// Returns true if formatless image load is supported.
+    bool IsFormatlessImageLoadSupported() const {
+        return is_formatless_image_load_supported;
+    }
+
     /// Returns true if the device supports VK_EXT_scalar_block_layout.
     bool IsKhrUniformBufferStandardLayoutSupported() const {
         return khr_uniform_buffer_standard_layout;
@@ -167,6 +167,11 @@ public:
         return ext_shader_viewport_index_layer;
     }
 
+    /// Returns true if the device supports VK_EXT_transform_feedback.
+    bool IsExtTransformFeedbackSupported() const {
+        return ext_transform_feedback;
+    }
+
     /// Returns true if the device supports VK_NV_device_diagnostic_checkpoints.
     bool IsNvDeviceDiagnosticCheckpoints() const {
         return nv_device_diagnostic_checkpoints;
@@ -214,26 +219,26 @@ private:
     static std::unordered_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.
-    vk::PhysicalDeviceProperties properties;   ///< Device properties.
-    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::DriverIdKHR driver_id{};               ///< Driver ID.
-    vk::ShaderStageFlags guest_warp_stages{};  ///< Stages where the guest warp size can be forced.
-    bool is_optimal_astc_supported{};          ///< Support for native ASTC.
-    bool is_float16_supported{};               ///< Support for float16 arithmetics.
-    bool is_warp_potentially_bigger{};         ///< Host warp size can be bigger than guest.
+    const vk::PhysicalDevice physical;        ///< Physical device.
+    vk::DispatchLoaderDynamic dld;            ///< Device function pointers.
+    vk::PhysicalDeviceProperties properties;  ///< Device properties.
+    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::DriverIdKHR driver_id{};              ///< Driver ID.
+    vk::ShaderStageFlags guest_warp_stages{}; ///< Stages where the guest warp size can be forced.ed
+    bool is_optimal_astc_supported{};         ///< Support for native ASTC.
+    bool is_float16_supported{};              ///< Support for float16 arithmetics.
+    bool is_warp_potentially_bigger{};        ///< Host warp size can be bigger than guest.
+    bool is_formatless_image_load_supported{}; ///< Support for shader image read without format.
     bool khr_uniform_buffer_standard_layout{}; ///< Support for std430 on UBOs.
     bool ext_index_type_uint8{};               ///< Support for VK_EXT_index_type_uint8.
     bool ext_depth_range_unrestricted{};       ///< Support for VK_EXT_depth_range_unrestricted.
     bool ext_shader_viewport_index_layer{};    ///< Support for VK_EXT_shader_viewport_index_layer.
+    bool ext_transform_feedback{};             ///< Support for VK_EXT_transform_feedback.
     bool nv_device_diagnostic_checkpoints{};   ///< Support for VK_NV_device_diagnostic_checkpoints.
-    bool is_shader_storage_img_read_without_format_supported{}; ///< Support for shader storage
-                                                                ///< image read without format
 
     // Telemetry parameters
     std::string vendor_name;                      ///< Device's driver name.
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
index b155dfb49..6a02403c1 100644
--- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
@@ -97,8 +97,7 @@ UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplat
     u32 offset = 0;
     for (const auto& stage : program) {
         if (stage) {
-            FillDescriptorUpdateTemplateEntries(device, stage->entries, binding, offset,
-                                                template_entries);
+            FillDescriptorUpdateTemplateEntries(stage->entries, binding, offset, template_entries);
         }
     }
     if (template_entries.empty()) {
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
index 7ddf7d3ee..557b9d662 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
@@ -36,6 +36,13 @@ using Tegra::Engines::ShaderType;
 
 namespace {
 
+// C++20's using enum
+constexpr auto eUniformBuffer = vk::DescriptorType::eUniformBuffer;
+constexpr auto eStorageBuffer = vk::DescriptorType::eStorageBuffer;
+constexpr auto eUniformTexelBuffer = vk::DescriptorType::eUniformTexelBuffer;
+constexpr auto eCombinedImageSampler = vk::DescriptorType::eCombinedImageSampler;
+constexpr auto eStorageImage = vk::DescriptorType::eStorageImage;
+
 constexpr VideoCommon::Shader::CompilerSettings compiler_settings{
     VideoCommon::Shader::CompileDepth::FullDecompile};
 
@@ -119,23 +126,32 @@ ShaderType GetShaderType(Maxwell::ShaderProgram program) {
     }
 }
 
+template <vk::DescriptorType descriptor_type, class Container>
+void AddBindings(std::vector<vk::DescriptorSetLayoutBinding>& bindings, u32& binding,
+                 vk::ShaderStageFlags stage_flags, const Container& container) {
+    const u32 num_entries = static_cast<u32>(std::size(container));
+    for (std::size_t i = 0; i < num_entries; ++i) {
+        u32 count = 1;
+        if constexpr (descriptor_type == eCombinedImageSampler) {
+            // Combined image samplers can be arrayed.
+            count = container[i].Size();
+        }
+        bindings.emplace_back(binding++, descriptor_type, count, stage_flags, nullptr);
+    }
+}
+
 u32 FillDescriptorLayout(const ShaderEntries& entries,
                          std::vector<vk::DescriptorSetLayoutBinding>& bindings,
                          Maxwell::ShaderProgram program_type, u32 base_binding) {
     const ShaderType stage = GetStageFromProgram(program_type);
-    const vk::ShaderStageFlags stage_flags = MaxwellToVK::ShaderStage(stage);
+    const vk::ShaderStageFlags flags = MaxwellToVK::ShaderStage(stage);
 
     u32 binding = base_binding;
-    const auto AddBindings = [&](vk::DescriptorType descriptor_type, std::size_t num_entries) {
-        for (std::size_t i = 0; i < num_entries; ++i) {
-            bindings.emplace_back(binding++, descriptor_type, 1, stage_flags, nullptr);
-        }
-    };
-    AddBindings(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
-    AddBindings(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
-    AddBindings(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
-    AddBindings(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
-    AddBindings(vk::DescriptorType::eStorageImage, entries.images.size());
+    AddBindings<eUniformBuffer>(bindings, binding, flags, entries.const_buffers);
+    AddBindings<eStorageBuffer>(bindings, binding, flags, entries.global_buffers);
+    AddBindings<eUniformTexelBuffer>(bindings, binding, flags, entries.texel_buffers);
+    AddBindings<eCombinedImageSampler>(bindings, binding, flags, entries.samplers);
+    AddBindings<eStorageImage>(bindings, binding, flags, entries.images);
     return binding;
 }
 
@@ -145,8 +161,8 @@ CachedShader::CachedShader(Core::System& system, Tegra::Engines::ShaderType stag
                            GPUVAddr gpu_addr, VAddr cpu_addr, u8* host_ptr,
                            ProgramCode program_code, u32 main_offset)
     : RasterizerCacheObject{host_ptr}, gpu_addr{gpu_addr}, cpu_addr{cpu_addr},
-      program_code{std::move(program_code)}, locker{stage, GetEngine(system, stage)},
-      shader_ir{this->program_code, main_offset, compiler_settings, locker},
+      program_code{std::move(program_code)}, registry{stage, GetEngine(system, stage)},
+      shader_ir{this->program_code, main_offset, compiler_settings, registry},
       entries{GenerateShaderEntries(shader_ir)} {}
 
 CachedShader::~CachedShader() = default;
@@ -163,24 +179,19 @@ Tegra::Engines::ConstBufferEngineInterface& CachedShader::GetEngine(
 VKPipelineCache::VKPipelineCache(Core::System& system, RasterizerVulkan& rasterizer,
                                  const VKDevice& device, VKScheduler& scheduler,
                                  VKDescriptorPool& descriptor_pool,
-                                 VKUpdateDescriptorQueue& update_descriptor_queue)
+                                 VKUpdateDescriptorQueue& update_descriptor_queue,
+                                 VKRenderPassCache& renderpass_cache)
     : RasterizerCache{rasterizer}, system{system}, device{device}, scheduler{scheduler},
       descriptor_pool{descriptor_pool}, update_descriptor_queue{update_descriptor_queue},
-      renderpass_cache(device) {}
+      renderpass_cache{renderpass_cache} {}
 
 VKPipelineCache::~VKPipelineCache() = default;
 
 std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
     const auto& gpu = system.GPU().Maxwell3D();
-    auto& dirty = system.GPU().Maxwell3D().dirty.shaders;
-    if (!dirty) {
-        return last_shaders;
-    }
-    dirty = false;
 
     std::array<Shader, Maxwell::MaxShaderProgram> shaders;
     for (std::size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
-        const auto& shader_config = gpu.regs.shader_config[index];
         const auto program{static_cast<Maxwell::ShaderProgram>(index)};
 
         // Skip stages that are not enabled
@@ -262,9 +273,9 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach
     specialization.workgroup_size = key.workgroup_size;
     specialization.shared_memory_size = key.shared_memory_size;
 
-    const SPIRVShader spirv_shader{
-        Decompile(device, shader->GetIR(), ShaderType::Compute, specialization),
-        shader->GetEntries()};
+    const SPIRVShader spirv_shader{Decompile(device, shader->GetIR(), ShaderType::Compute,
+                                             shader->GetRegistry(), specialization),
+                                   shader->GetEntries()};
     entry = std::make_unique<VKComputePipeline>(device, scheduler, descriptor_pool,
                                                 update_descriptor_queue, spirv_shader);
     return *entry;
@@ -313,8 +324,7 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
     const auto& gpu = system.GPU().Maxwell3D();
 
     Specialization specialization;
-    specialization.primitive_topology = fixed_state.input_assembly.topology;
-    if (specialization.primitive_topology == Maxwell::PrimitiveTopology::Points) {
+    if (fixed_state.input_assembly.topology == Maxwell::PrimitiveTopology::Points) {
         ASSERT(fixed_state.input_assembly.point_size != 0.0f);
         specialization.point_size = fixed_state.input_assembly.point_size;
     }
@@ -322,9 +332,6 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
         specialization.attribute_types[i] = fixed_state.vertex_input.attributes[i].type;
     }
     specialization.ndc_minus_one_to_one = fixed_state.rasterizer.ndc_minus_one_to_one;
-    specialization.tessellation.primitive = fixed_state.tessellation.primitive;
-    specialization.tessellation.spacing = fixed_state.tessellation.spacing;
-    specialization.tessellation.clockwise = fixed_state.tessellation.clockwise;
 
     SPIRVProgram program;
     std::vector<vk::DescriptorSetLayoutBinding> bindings;
@@ -345,8 +352,9 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
         const std::size_t stage = index == 0 ? 0 : index - 1; // Stage indices are 0 - 5
         const auto program_type = GetShaderType(program_enum);
         const auto& entries = shader->GetEntries();
-        program[stage] = {Decompile(device, shader->GetIR(), program_type, specialization),
-                          entries};
+        program[stage] = {
+            Decompile(device, shader->GetIR(), program_type, shader->GetRegistry(), specialization),
+            entries};
 
         if (program_enum == Maxwell::ShaderProgram::VertexA) {
             // VertexB was combined with VertexA, so we skip the VertexB iteration
@@ -361,32 +369,45 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
     return {std::move(program), std::move(bindings)};
 }
 
-void FillDescriptorUpdateTemplateEntries(
-    const VKDevice& device, const ShaderEntries& entries, u32& binding, u32& offset,
-    std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) {
-    static constexpr auto entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry));
-    const auto AddEntry = [&](vk::DescriptorType descriptor_type, std::size_t count_) {
-        const u32 count = static_cast<u32>(count_);
-        if (descriptor_type == vk::DescriptorType::eUniformTexelBuffer &&
-            device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) {
-            // Nvidia has a bug where updating multiple uniform texels at once causes the driver to
-            // crash.
-            for (u32 i = 0; i < count; ++i) {
-                template_entries.emplace_back(binding + i, 0, 1, descriptor_type,
-                                              offset + i * entry_size, entry_size);
-            }
-        } else if (count != 0) {
-            template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size);
+template <vk::DescriptorType descriptor_type, class Container>
+void AddEntry(std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries, u32& binding,
+              u32& offset, const Container& container) {
+    static constexpr u32 entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry));
+    const u32 count = static_cast<u32>(std::size(container));
+
+    if constexpr (descriptor_type == eCombinedImageSampler) {
+        for (u32 i = 0; i < count; ++i) {
+            const u32 num_samplers = container[i].Size();
+            template_entries.emplace_back(binding, 0, num_samplers, descriptor_type, offset,
+                                          entry_size);
+            ++binding;
+            offset += num_samplers * entry_size;
         }
-        offset += count * entry_size;
-        binding += count;
-    };
+        return;
+    }
 
-    AddEntry(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
-    AddEntry(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
-    AddEntry(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
-    AddEntry(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
-    AddEntry(vk::DescriptorType::eStorageImage, entries.images.size());
+    if constexpr (descriptor_type == eUniformTexelBuffer) {
+        // Nvidia has a bug where updating multiple uniform texels at once causes the driver to
+        // crash.
+        for (u32 i = 0; i < count; ++i) {
+            template_entries.emplace_back(binding + i, 0, 1, descriptor_type,
+                                          offset + i * entry_size, entry_size);
+        }
+    } else if (count > 0) {
+        template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size);
+    }
+    offset += count * entry_size;
+    binding += count;
+}
+
+void FillDescriptorUpdateTemplateEntries(
+    const ShaderEntries& entries, u32& binding, u32& offset,
+    std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) {
+    AddEntry<eUniformBuffer>(template_entries, offset, binding, entries.const_buffers);
+    AddEntry<eStorageBuffer>(template_entries, offset, binding, entries.global_buffers);
+    AddEntry<eUniformTexelBuffer>(template_entries, offset, binding, entries.texel_buffers);
+    AddEntry<eCombinedImageSampler>(template_entries, offset, binding, entries.samplers);
+    AddEntry<eStorageImage>(template_entries, offset, binding, entries.images);
 }
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
index 8678fc9c3..c4c112290 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
@@ -25,7 +25,7 @@
 #include "video_core/renderer_vulkan/vk_renderpass_cache.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_shader_decompiler.h"
-#include "video_core/shader/const_buffer_locker.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 #include "video_core/surface.h"
 
@@ -132,6 +132,10 @@ public:
         return shader_ir;
     }
 
+    const VideoCommon::Shader::Registry& GetRegistry() const {
+        return registry;
+    }
+
     const VideoCommon::Shader::ShaderIR& GetIR() const {
         return shader_ir;
     }
@@ -147,7 +151,7 @@ private:
     GPUVAddr gpu_addr{};
     VAddr cpu_addr{};
     ProgramCode program_code;
-    VideoCommon::Shader::ConstBufferLocker locker;
+    VideoCommon::Shader::Registry registry;
     VideoCommon::Shader::ShaderIR shader_ir;
     ShaderEntries entries;
 };
@@ -157,7 +161,8 @@ public:
     explicit VKPipelineCache(Core::System& system, RasterizerVulkan& rasterizer,
                              const VKDevice& device, VKScheduler& scheduler,
                              VKDescriptorPool& descriptor_pool,
-                             VKUpdateDescriptorQueue& update_descriptor_queue);
+                             VKUpdateDescriptorQueue& update_descriptor_queue,
+                             VKRenderPassCache& renderpass_cache);
     ~VKPipelineCache();
 
     std::array<Shader, Maxwell::MaxShaderProgram> GetShaders();
@@ -180,8 +185,7 @@ private:
     VKScheduler& scheduler;
     VKDescriptorPool& descriptor_pool;
     VKUpdateDescriptorQueue& update_descriptor_queue;
-
-    VKRenderPassCache renderpass_cache;
+    VKRenderPassCache& renderpass_cache;
 
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 
@@ -194,7 +198,7 @@ private:
 };
 
 void FillDescriptorUpdateTemplateEntries(
-    const VKDevice& device, const ShaderEntries& entries, u32& binding, u32& offset,
+    const ShaderEntries& entries, u32& binding, u32& offset,
     std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries);
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index 31c078f6a..58c69b786 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -36,6 +36,7 @@
 #include "video_core/renderer_vulkan/vk_sampler_cache.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
 #include "video_core/renderer_vulkan/vk_texture_cache.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
 
@@ -105,17 +106,20 @@ void TransitionImages(const std::vector<ImageView>& views, vk::PipelineStageFlag
 
 template <typename Engine, typename Entry>
 Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
-                                               std::size_t stage) {
+                                               std::size_t stage, std::size_t index = 0) {
     const auto stage_type = static_cast<Tegra::Engines::ShaderType>(stage);
     if (entry.IsBindless()) {
         const Tegra::Texture::TextureHandle tex_handle =
             engine.AccessConstBuffer32(stage_type, entry.GetBuffer(), entry.GetOffset());
         return engine.GetTextureInfo(tex_handle);
     }
+    const auto& gpu_profile = engine.AccessGuestDriverProfile();
+    const u32 entry_offset = static_cast<u32>(index * gpu_profile.GetTextureHandlerSize());
+    const u32 offset = entry.GetOffset() + entry_offset;
     if constexpr (std::is_same_v<Engine, Tegra::Engines::Maxwell3D>) {
-        return engine.GetStageTexture(stage_type, entry.GetOffset());
+        return engine.GetStageTexture(stage_type, offset);
     } else {
-        return engine.GetTexture(entry.GetOffset());
+        return engine.GetTexture(offset);
     }
 }
 
@@ -277,17 +281,19 @@ void RasterizerVulkan::DrawParameters::Draw(vk::CommandBuffer cmdbuf,
 RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWindow& renderer,
                                    VKScreenInfo& screen_info, const VKDevice& device,
                                    VKResourceManager& resource_manager,
-                                   VKMemoryManager& memory_manager, VKScheduler& scheduler)
+                                   VKMemoryManager& memory_manager, StateTracker& state_tracker,
+                                   VKScheduler& scheduler)
     : RasterizerAccelerated{system.Memory()}, system{system}, render_window{renderer},
       screen_info{screen_info}, device{device}, resource_manager{resource_manager},
-      memory_manager{memory_manager}, scheduler{scheduler},
+      memory_manager{memory_manager}, state_tracker{state_tracker}, scheduler{scheduler},
       staging_pool(device, memory_manager, scheduler), descriptor_pool(device),
-      update_descriptor_queue(device, scheduler),
+      update_descriptor_queue(device, scheduler), renderpass_cache(device),
       quad_array_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
       uint8_pass(device, scheduler, descriptor_pool, staging_pool, update_descriptor_queue),
       texture_cache(system, *this, device, resource_manager, memory_manager, scheduler,
                     staging_pool),
-      pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue),
+      pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue,
+                     renderpass_cache),
       buffer_cache(*this, system, device, memory_manager, scheduler, staging_pool),
       sampler_cache(device), query_cache(system, *this, device, scheduler) {
     scheduler.SetQueryCache(query_cache);
@@ -342,6 +348,8 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
             [&pipeline](auto cmdbuf, auto& dld) { cmdbuf.setCheckpointNV(&pipeline, dld); });
     }
 
+    BeginTransformFeedback();
+
     const auto pipeline_layout = pipeline.GetLayout();
     const auto descriptor_set = pipeline.CommitDescriptorSet();
     scheduler.Record([pipeline_layout, descriptor_set, draw_params](auto cmdbuf, auto& dld) {
@@ -351,18 +359,23 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
         }
         draw_params.Draw(cmdbuf, dld);
     });
+
+    EndTransformFeedback();
 }
 
 void RasterizerVulkan::Clear() {
     MICROPROFILE_SCOPE(Vulkan_Clearing);
 
-    query_cache.UpdateCounters();
-
     const auto& gpu = system.GPU().Maxwell3D();
     if (!system.GPU().Maxwell3D().ShouldExecute()) {
         return;
     }
 
+    sampled_views.clear();
+    image_views.clear();
+
+    query_cache.UpdateCounters();
+
     const auto& regs = gpu.regs;
     const bool use_color = regs.clear_buffers.R || regs.clear_buffers.G || regs.clear_buffers.B ||
                            regs.clear_buffers.A;
@@ -371,52 +384,54 @@ void RasterizerVulkan::Clear() {
     if (!use_color && !use_depth && !use_stencil) {
         return;
     }
-    // Clearing images requires to be out of a renderpass
-    scheduler.RequestOutsideRenderPassOperationContext();
 
-    // TODO(Rodrigo): Implement clears rendering a quad or using beginning a renderpass.
+    [[maybe_unused]] const auto texceptions = UpdateAttachments();
+    DEBUG_ASSERT(texceptions.none());
+    SetupImageTransitions(0, color_attachments, zeta_attachment);
 
-    if (use_color) {
-        View color_view;
-        {
-            MICROPROFILE_SCOPE(Vulkan_RenderTargets);
-            color_view = texture_cache.GetColorBufferSurface(regs.clear_buffers.RT.Value(), false);
-        }
+    const vk::RenderPass renderpass = renderpass_cache.GetRenderPass(GetRenderPassParams(0));
+    const auto [framebuffer, render_area] = ConfigureFramebuffers(renderpass);
+    scheduler.RequestRenderpass({renderpass, framebuffer, {{0, 0}, render_area}, 0, nullptr});
 
-        color_view->Transition(vk::ImageLayout::eTransferDstOptimal,
-                               vk::PipelineStageFlagBits::eTransfer,
-                               vk::AccessFlagBits::eTransferWrite);
+    const auto& scissor = regs.scissor_test[0];
+    const vk::Offset2D scissor_offset(scissor.min_x, scissor.min_y);
+    vk::Extent2D scissor_extent{scissor.max_x - scissor.min_x, scissor.max_y - scissor.min_y};
+    scissor_extent.width = std::min(scissor_extent.width, render_area.width);
+    scissor_extent.height = std::min(scissor_extent.height, render_area.height);
 
+    const u32 layer = regs.clear_buffers.layer;
+    const vk::ClearRect clear_rect({scissor_offset, scissor_extent}, layer, 1);
+
+    if (use_color) {
         const std::array clear_color = {regs.clear_color[0], regs.clear_color[1],
                                         regs.clear_color[2], regs.clear_color[3]};
-        const vk::ClearColorValue clear(clear_color);
-        scheduler.Record([image = color_view->GetImage(),
-                          subresource = color_view->GetImageSubresourceRange(),
-                          clear](auto cmdbuf, auto& dld) {
-            cmdbuf.clearColorImage(image, vk::ImageLayout::eTransferDstOptimal, clear, subresource,
-                                   dld);
+        const vk::ClearValue clear_value{clear_color};
+        const u32 color_attachment = regs.clear_buffers.RT;
+        scheduler.Record([color_attachment, clear_value, clear_rect](auto cmdbuf, auto& dld) {
+            const vk::ClearAttachment attachment(vk::ImageAspectFlagBits::eColor, color_attachment,
+                                                 clear_value);
+            cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld);
         });
     }
-    if (use_depth || use_stencil) {
-        View zeta_surface;
-        {
-            MICROPROFILE_SCOPE(Vulkan_RenderTargets);
-            zeta_surface = texture_cache.GetDepthBufferSurface(false);
-        }
 
-        zeta_surface->Transition(vk::ImageLayout::eTransferDstOptimal,
-                                 vk::PipelineStageFlagBits::eTransfer,
-                                 vk::AccessFlagBits::eTransferWrite);
-
-        const vk::ClearDepthStencilValue clear(regs.clear_depth,
-                                               static_cast<u32>(regs.clear_stencil));
-        scheduler.Record([image = zeta_surface->GetImage(),
-                          subresource = zeta_surface->GetImageSubresourceRange(),
-                          clear](auto cmdbuf, auto& dld) {
-            cmdbuf.clearDepthStencilImage(image, vk::ImageLayout::eTransferDstOptimal, clear,
-                                          subresource, dld);
-        });
+    if (!use_depth && !use_stencil) {
+        return;
+    }
+    vk::ImageAspectFlags aspect_flags;
+    if (use_depth) {
+        aspect_flags |= vk::ImageAspectFlagBits::eDepth;
+    }
+    if (use_stencil) {
+        aspect_flags |= vk::ImageAspectFlagBits::eStencil;
     }
+
+    scheduler.Record([clear_depth = regs.clear_depth, clear_stencil = regs.clear_stencil,
+                      clear_rect, aspect_flags](auto cmdbuf, auto& dld) {
+        const vk::ClearDepthStencilValue clear_zeta(clear_depth, clear_stencil);
+        const vk::ClearValue clear_value{clear_zeta};
+        const vk::ClearAttachment attachment(aspect_flags, 0, clear_value);
+        cmdbuf.clearAttachments(1, &attachment, 1, &clear_rect, dld);
+    });
 }
 
 void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
@@ -533,8 +548,6 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
 
     // Verify that the cached surface is the same size and format as the requested framebuffer
     const auto& params{surface->GetSurfaceParams()};
-    const auto& pixel_format{
-        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");
 
@@ -545,6 +558,10 @@ bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
     return true;
 }
 
+void RasterizerVulkan::SetupDirtyFlags() {
+    state_tracker.Initialize();
+}
+
 void RasterizerVulkan::FlushWork() {
     static constexpr u32 DRAWS_TO_DISPATCH = 4096;
 
@@ -568,9 +585,9 @@ void RasterizerVulkan::FlushWork() {
 
 RasterizerVulkan::Texceptions RasterizerVulkan::UpdateAttachments() {
     MICROPROFILE_SCOPE(Vulkan_RenderTargets);
-    auto& dirty = system.GPU().Maxwell3D().dirty;
-    const bool update_rendertargets = dirty.render_settings;
-    dirty.render_settings = false;
+    auto& dirty = system.GPU().Maxwell3D().dirty.flags;
+    const bool update_rendertargets = dirty[VideoCommon::Dirty::RenderTargets];
+    dirty[VideoCommon::Dirty::RenderTargets] = false;
 
     texture_cache.GuardRenderTargets(true);
 
@@ -611,33 +628,34 @@ bool RasterizerVulkan::WalkAttachmentOverlaps(const CachedSurfaceView& attachmen
 std::tuple<vk::Framebuffer, vk::Extent2D> RasterizerVulkan::ConfigureFramebuffers(
     vk::RenderPass renderpass) {
     FramebufferCacheKey key{renderpass, std::numeric_limits<u32>::max(),
-                            std::numeric_limits<u32>::max()};
+                            std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()};
 
-    const auto MarkAsModifiedAndPush = [&](const View& view) {
-        if (view == nullptr) {
+    const auto try_push = [&](const View& view) {
+        if (!view) {
             return false;
         }
         key.views.push_back(view->GetHandle());
         key.width = std::min(key.width, view->GetWidth());
         key.height = std::min(key.height, view->GetHeight());
+        key.layers = std::min(key.layers, view->GetNumLayers());
         return true;
     };
 
     for (std::size_t index = 0; index < std::size(color_attachments); ++index) {
-        if (MarkAsModifiedAndPush(color_attachments[index])) {
+        if (try_push(color_attachments[index])) {
             texture_cache.MarkColorBufferInUse(index);
         }
     }
-    if (MarkAsModifiedAndPush(zeta_attachment)) {
+    if (try_push(zeta_attachment)) {
         texture_cache.MarkDepthBufferInUse();
     }
 
     const auto [fbentry, is_cache_miss] = framebuffer_cache.try_emplace(key);
     auto& framebuffer = fbentry->second;
     if (is_cache_miss) {
-        const vk::FramebufferCreateInfo framebuffer_ci({}, key.renderpass,
-                                                       static_cast<u32>(key.views.size()),
-                                                       key.views.data(), key.width, key.height, 1);
+        const vk::FramebufferCreateInfo framebuffer_ci(
+            {}, key.renderpass, static_cast<u32>(key.views.size()), key.views.data(), key.width,
+            key.height, key.layers);
         const auto dev = device.GetLogical();
         const auto& dld = device.GetDispatchLoader();
         framebuffer = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld);
@@ -719,13 +737,51 @@ void RasterizerVulkan::SetupImageTransitions(
 }
 
 void RasterizerVulkan::UpdateDynamicStates() {
-    auto& gpu = system.GPU().Maxwell3D();
-    UpdateViewportsState(gpu);
-    UpdateScissorsState(gpu);
-    UpdateDepthBias(gpu);
-    UpdateBlendConstants(gpu);
-    UpdateDepthBounds(gpu);
-    UpdateStencilFaces(gpu);
+    auto& regs = system.GPU().Maxwell3D().regs;
+    UpdateViewportsState(regs);
+    UpdateScissorsState(regs);
+    UpdateDepthBias(regs);
+    UpdateBlendConstants(regs);
+    UpdateDepthBounds(regs);
+    UpdateStencilFaces(regs);
+}
+
+void RasterizerVulkan::BeginTransformFeedback() {
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    if (regs.tfb_enabled == 0) {
+        return;
+    }
+
+    UNIMPLEMENTED_IF(regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::TesselationControl) ||
+                     regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::TesselationEval) ||
+                     regs.IsShaderConfigEnabled(Maxwell::ShaderProgram::Geometry));
+
+    UNIMPLEMENTED_IF(regs.tfb_bindings[1].buffer_enable);
+    UNIMPLEMENTED_IF(regs.tfb_bindings[2].buffer_enable);
+    UNIMPLEMENTED_IF(regs.tfb_bindings[3].buffer_enable);
+
+    const auto& binding = regs.tfb_bindings[0];
+    UNIMPLEMENTED_IF(binding.buffer_enable == 0);
+    UNIMPLEMENTED_IF(binding.buffer_offset != 0);
+
+    const GPUVAddr gpu_addr = binding.Address();
+    const std::size_t size = binding.buffer_size;
+    const auto [buffer, offset] = buffer_cache.UploadMemory(gpu_addr, size, 4, true);
+
+    scheduler.Record([buffer = *buffer, offset = offset, size](auto cmdbuf, auto& dld) {
+        cmdbuf.bindTransformFeedbackBuffersEXT(0, {buffer}, {offset}, {size}, dld);
+        cmdbuf.beginTransformFeedbackEXT(0, {}, {}, dld);
+    });
+}
+
+void RasterizerVulkan::EndTransformFeedback() {
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    if (regs.tfb_enabled == 0) {
+        return;
+    }
+
+    scheduler.Record(
+        [](auto cmdbuf, auto& dld) { cmdbuf.endTransformFeedbackEXT(0, {}, {}, dld); });
 }
 
 void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex_input,
@@ -835,14 +891,16 @@ void RasterizerVulkan::SetupGraphicsTextures(const ShaderEntries& entries, std::
     MICROPROFILE_SCOPE(Vulkan_Textures);
     const auto& gpu = system.GPU().Maxwell3D();
     for (const auto& entry : entries.samplers) {
-        const auto texture = GetTextureInfo(gpu, entry, stage);
-        SetupTexture(texture, entry);
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
+            const auto texture = GetTextureInfo(gpu, entry, stage, i);
+            SetupTexture(texture, entry);
+        }
     }
 }
 
 void RasterizerVulkan::SetupGraphicsImages(const ShaderEntries& entries, std::size_t stage) {
     MICROPROFILE_SCOPE(Vulkan_Images);
-    const auto& gpu = system.GPU().KeplerCompute();
+    const auto& gpu = system.GPU().Maxwell3D();
     for (const auto& entry : entries.images) {
         const auto tic = GetTextureInfo(gpu, entry, stage).tic;
         SetupImage(tic, entry);
@@ -885,8 +943,10 @@ void RasterizerVulkan::SetupComputeTextures(const ShaderEntries& entries) {
     MICROPROFILE_SCOPE(Vulkan_Textures);
     const auto& gpu = system.GPU().KeplerCompute();
     for (const auto& entry : entries.samplers) {
-        const auto texture = GetTextureInfo(gpu, entry, ComputeShaderIndex);
-        SetupTexture(texture, entry);
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
+            const auto texture = GetTextureInfo(gpu, entry, ComputeShaderIndex, i);
+            SetupTexture(texture, entry);
+        }
     }
 }
 
@@ -901,6 +961,13 @@ void RasterizerVulkan::SetupComputeImages(const ShaderEntries& entries) {
 
 void RasterizerVulkan::SetupConstBuffer(const ConstBufferEntry& entry,
                                         const Tegra::Engines::ConstBufferInfo& buffer) {
+    if (!buffer.enabled) {
+        // Set values to zero to unbind buffers
+        update_descriptor_queue.AddBuffer(buffer_cache.GetEmptyBuffer(sizeof(float)), 0,
+                                          sizeof(float));
+        return;
+    }
+
     // Align the size to avoid bad std140 interactions
     const std::size_t size =
         Common::AlignUp(CalculateConstBufferSize(entry, buffer), 4 * sizeof(float));
@@ -971,12 +1038,10 @@ void RasterizerVulkan::SetupImage(const Tegra::Texture::TICEntry& tic, const Ima
     image_views.push_back(ImageView{std::move(view), image_layout});
 }
 
-void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.viewport_transform && scheduler.TouchViewports()) {
+void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchViewports()) {
         return;
     }
-    gpu.dirty.viewport_transform = false;
-    const auto& regs = gpu.regs;
     const std::array viewports{
         GetViewportState(device, regs, 0),  GetViewportState(device, regs, 1),
         GetViewportState(device, regs, 2),  GetViewportState(device, regs, 3),
@@ -991,12 +1056,10 @@ void RasterizerVulkan::UpdateViewportsState(Tegra::Engines::Maxwell3D& gpu) {
     });
 }
 
-void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.scissor_test && scheduler.TouchScissors()) {
+void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchScissors()) {
         return;
     }
-    gpu.dirty.scissor_test = false;
-    const auto& regs = gpu.regs;
     const std::array scissors = {
         GetScissorState(regs, 0),  GetScissorState(regs, 1),  GetScissorState(regs, 2),
         GetScissorState(regs, 3),  GetScissorState(regs, 4),  GetScissorState(regs, 5),
@@ -1009,46 +1072,39 @@ void RasterizerVulkan::UpdateScissorsState(Tegra::Engines::Maxwell3D& gpu) {
     });
 }
 
-void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.polygon_offset && scheduler.TouchDepthBias()) {
+void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchDepthBias()) {
         return;
     }
-    gpu.dirty.polygon_offset = false;
-    const auto& regs = gpu.regs;
     scheduler.Record([constant = regs.polygon_offset_units, clamp = regs.polygon_offset_clamp,
                       factor = regs.polygon_offset_factor](auto cmdbuf, auto& dld) {
         cmdbuf.setDepthBias(constant, clamp, factor / 2.0f, dld);
     });
 }
 
-void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.blend_state && scheduler.TouchBlendConstants()) {
+void RasterizerVulkan::UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchBlendConstants()) {
         return;
     }
-    gpu.dirty.blend_state = false;
-    const std::array blend_color = {gpu.regs.blend_color.r, gpu.regs.blend_color.g,
-                                    gpu.regs.blend_color.b, gpu.regs.blend_color.a};
+    const std::array blend_color = {regs.blend_color.r, regs.blend_color.g, regs.blend_color.b,
+                                    regs.blend_color.a};
     scheduler.Record([blend_color](auto cmdbuf, auto& dld) {
         cmdbuf.setBlendConstants(blend_color.data(), dld);
     });
 }
 
-void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.depth_bounds_values && scheduler.TouchDepthBounds()) {
+void RasterizerVulkan::UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchDepthBounds()) {
         return;
     }
-    gpu.dirty.depth_bounds_values = false;
-    const auto& regs = gpu.regs;
     scheduler.Record([min = regs.depth_bounds[0], max = regs.depth_bounds[1]](
                          auto cmdbuf, auto& dld) { cmdbuf.setDepthBounds(min, max, dld); });
 }
 
-void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D& gpu) {
-    if (!gpu.dirty.stencil_test && scheduler.TouchStencilValues()) {
+void RasterizerVulkan::UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs) {
+    if (!state_tracker.TouchStencilProperties()) {
         return;
     }
-    gpu.dirty.stencil_test = false;
-    const auto& regs = gpu.regs;
     if (regs.stencil_two_side_enable) {
         // Separate values per face
         scheduler.Record(
@@ -1099,7 +1155,7 @@ std::size_t RasterizerVulkan::CalculateVertexArraysSize() const {
         // This implementation assumes that all attributes are used in the shader.
         const GPUVAddr start{regs.vertex_array[index].StartAddress()};
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
-        DEBUG_ASSERT(end > start);
+        DEBUG_ASSERT(end >= start);
 
         size += (end - start + 1) * regs.vertex_array[index].enable;
     }
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index 138903d60..3185868e9 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -56,6 +56,7 @@ struct FramebufferCacheKey {
     vk::RenderPass renderpass{};
     u32 width = 0;
     u32 height = 0;
+    u32 layers = 0;
     ImageViewsPack views;
 
     std::size_t Hash() const noexcept {
@@ -66,12 +67,17 @@ struct FramebufferCacheKey {
         }
         boost::hash_combine(hash, width);
         boost::hash_combine(hash, height);
+        boost::hash_combine(hash, layers);
         return hash;
     }
 
     bool operator==(const FramebufferCacheKey& rhs) const noexcept {
-        return std::tie(renderpass, views, width, height) ==
-               std::tie(rhs.renderpass, rhs.views, rhs.width, rhs.height);
+        return std::tie(renderpass, views, width, height, layers) ==
+               std::tie(rhs.renderpass, rhs.views, rhs.width, rhs.height, rhs.layers);
+    }
+
+    bool operator!=(const FramebufferCacheKey& rhs) const noexcept {
+        return !operator==(rhs);
     }
 };
 
@@ -90,6 +96,7 @@ struct hash<Vulkan::FramebufferCacheKey> {
 
 namespace Vulkan {
 
+class StateTracker;
 class BufferBindings;
 
 struct ImageView {
@@ -102,7 +109,7 @@ public:
     explicit RasterizerVulkan(Core::System& system, Core::Frontend::EmuWindow& render_window,
                               VKScreenInfo& screen_info, const VKDevice& device,
                               VKResourceManager& resource_manager, VKMemoryManager& memory_manager,
-                              VKScheduler& scheduler);
+                              StateTracker& state_tracker, VKScheduler& scheduler);
     ~RasterizerVulkan() override;
 
     void Draw(bool is_indexed, bool is_instanced) override;
@@ -121,6 +128,7 @@ public:
                                const Tegra::Engines::Fermi2D::Config& copy_config) override;
     bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
                            u32 pixel_stride) override;
+    void SetupDirtyFlags() override;
 
     /// Maximum supported size that a constbuffer can have in bytes.
     static constexpr std::size_t MaxConstbufferSize = 0x10000;
@@ -161,6 +169,10 @@ private:
 
     void UpdateDynamicStates();
 
+    void BeginTransformFeedback();
+
+    void EndTransformFeedback();
+
     bool WalkAttachmentOverlaps(const CachedSurfaceView& attachment);
 
     void SetupVertexArrays(FixedPipelineState::VertexInput& vertex_input,
@@ -209,12 +221,12 @@ private:
 
     void SetupImage(const Tegra::Texture::TICEntry& tic, const ImageEntry& entry);
 
-    void UpdateViewportsState(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateScissorsState(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateDepthBias(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateBlendConstants(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateDepthBounds(Tegra::Engines::Maxwell3D& gpu);
-    void UpdateStencilFaces(Tegra::Engines::Maxwell3D& gpu);
+    void UpdateViewportsState(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateScissorsState(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateBlendConstants(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateDepthBounds(Tegra::Engines::Maxwell3D::Regs& regs);
+    void UpdateStencilFaces(Tegra::Engines::Maxwell3D::Regs& regs);
 
     std::size_t CalculateGraphicsStreamBufferSize(bool is_indexed) const;
 
@@ -235,11 +247,13 @@ private:
     const VKDevice& device;
     VKResourceManager& resource_manager;
     VKMemoryManager& memory_manager;
+    StateTracker& state_tracker;
     VKScheduler& scheduler;
 
     VKStagingBufferPool staging_pool;
     VKDescriptorPool descriptor_pool;
     VKUpdateDescriptorQueue update_descriptor_queue;
+    VKRenderPassCache renderpass_cache;
     QuadArrayPass quad_array_pass;
     Uint8Pass uint8_pass;
 
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp
index 92bd6c344..b61d4fe63 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.cpp
+++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp
@@ -2,6 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <thread>
+#include <utility>
+
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "video_core/renderer_vulkan/declarations.h"
@@ -9,6 +15,7 @@
 #include "video_core/renderer_vulkan/vk_query_cache.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
 
 namespace Vulkan {
 
@@ -29,9 +36,10 @@ void VKScheduler::CommandChunk::ExecuteAll(vk::CommandBuffer cmdbuf,
     last = nullptr;
 }
 
-VKScheduler::VKScheduler(const VKDevice& device, VKResourceManager& resource_manager)
-    : device{device}, resource_manager{resource_manager}, next_fence{
-                                                              &resource_manager.CommitFence()} {
+VKScheduler::VKScheduler(const VKDevice& device, VKResourceManager& resource_manager,
+                         StateTracker& state_tracker)
+    : device{device}, resource_manager{resource_manager}, state_tracker{state_tracker},
+      next_fence{&resource_manager.CommitFence()} {
     AcquireNewChunk();
     AllocateNewContext();
     worker_thread = std::thread(&VKScheduler::WorkerThread, this);
@@ -157,12 +165,7 @@ void VKScheduler::AllocateNewContext() {
 
 void VKScheduler::InvalidateState() {
     state.graphics_pipeline = nullptr;
-    state.viewports = false;
-    state.scissors = false;
-    state.depth_bias = false;
-    state.blend_constants = false;
-    state.depth_bounds = false;
-    state.stencil_values = false;
+    state_tracker.InvalidateCommandBufferState();
 }
 
 void VKScheduler::EndPendingOperations() {
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h
index 62fd7858b..c7cc291c3 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.h
+++ b/src/video_core/renderer_vulkan/vk_scheduler.h
@@ -17,6 +17,7 @@
 
 namespace Vulkan {
 
+class StateTracker;
 class VKDevice;
 class VKFence;
 class VKQueryCache;
@@ -43,7 +44,8 @@ private:
 /// OpenGL-like operations on Vulkan command buffers.
 class VKScheduler {
 public:
-    explicit VKScheduler(const VKDevice& device, VKResourceManager& resource_manager);
+    explicit VKScheduler(const VKDevice& device, VKResourceManager& resource_manager,
+                         StateTracker& state_tracker);
     ~VKScheduler();
 
     /// Sends the current execution context to the GPU.
@@ -74,36 +76,6 @@ public:
         query_cache = &query_cache_;
     }
 
-    /// Returns true when viewports have been set in the current command buffer.
-    bool TouchViewports() {
-        return std::exchange(state.viewports, true);
-    }
-
-    /// Returns true when scissors have been set in the current command buffer.
-    bool TouchScissors() {
-        return std::exchange(state.scissors, true);
-    }
-
-    /// Returns true when depth bias have been set in the current command buffer.
-    bool TouchDepthBias() {
-        return std::exchange(state.depth_bias, true);
-    }
-
-    /// Returns true when blend constants have been set in the current command buffer.
-    bool TouchBlendConstants() {
-        return std::exchange(state.blend_constants, true);
-    }
-
-    /// Returns true when depth bounds have been set in the current command buffer.
-    bool TouchDepthBounds() {
-        return std::exchange(state.depth_bounds, true);
-    }
-
-    /// Returns true when stencil values have been set in the current command buffer.
-    bool TouchStencilValues() {
-        return std::exchange(state.stencil_values, true);
-    }
-
     /// Send work to a separate thread.
     template <typename T>
     void Record(T&& command) {
@@ -217,6 +189,8 @@ private:
 
     const VKDevice& device;
     VKResourceManager& resource_manager;
+    StateTracker& state_tracker;
+
     VKQueryCache* query_cache = nullptr;
 
     vk::CommandBuffer current_cmdbuf;
@@ -226,12 +200,6 @@ private:
     struct State {
         std::optional<vk::RenderPassBeginInfo> renderpass;
         vk::Pipeline graphics_pipeline;
-        bool viewports = false;
-        bool scissors = false;
-        bool depth_bias = false;
-        bool blend_constants = false;
-        bool depth_bounds = false;
-        bool stencil_values = false;
     } state;
 
     std::unique_ptr<CommandChunk> chunk;
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 2da622d15..51ecb5567 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -5,7 +5,9 @@
 #include <functional>
 #include <limits>
 #include <map>
+#include <optional>
 #include <type_traits>
+#include <unordered_map>
 #include <utility>
 
 #include <fmt/format.h>
@@ -24,6 +26,7 @@
 #include "video_core/renderer_vulkan/vk_shader_decompiler.h"
 #include "video_core/shader/node.h"
 #include "video_core/shader/shader_ir.h"
+#include "video_core/shader/transform_feedback.h"
 
 namespace Vulkan {
 
@@ -69,8 +72,9 @@ struct TexelBuffer {
 
 struct SampledImage {
     Id image_type{};
-    Id sampled_image_type{};
-    Id sampler{};
+    Id sampler_type{};
+    Id sampler_pointer_type{};
+    Id variable{};
 };
 
 struct StorageImage {
@@ -92,6 +96,12 @@ struct VertexIndices {
     std::optional<u32> clip_distances;
 };
 
+struct GenericVaryingDescription {
+    Id id = nullptr;
+    u32 first_element = 0;
+    bool is_scalar = false;
+};
+
 spv::Dim GetSamplerDim(const Sampler& sampler) {
     ASSERT(!sampler.IsBuffer());
     switch (sampler.GetType()) {
@@ -265,9 +275,13 @@ bool IsPrecise(Operation operand) {
 class SPIRVDecompiler final : public Sirit::Module {
 public:
     explicit SPIRVDecompiler(const VKDevice& device, const ShaderIR& ir, ShaderType stage,
-                             const Specialization& specialization)
+                             const Registry& registry, const Specialization& specialization)
         : Module(0x00010300), device{device}, ir{ir}, stage{stage}, header{ir.GetHeader()},
-          specialization{specialization} {
+          registry{registry}, specialization{specialization} {
+        if (stage != ShaderType::Compute) {
+            transform_feedback = BuildTransformFeedback(registry.GetGraphicsInfo());
+        }
+
         AddCapability(spv::Capability::Shader);
         AddCapability(spv::Capability::UniformAndStorageBuffer16BitAccess);
         AddCapability(spv::Capability::ImageQuery);
@@ -285,6 +299,15 @@ public:
         AddExtension("SPV_KHR_variable_pointers");
         AddExtension("SPV_KHR_shader_draw_parameters");
 
+        if (!transform_feedback.empty()) {
+            if (device.IsExtTransformFeedbackSupported()) {
+                AddCapability(spv::Capability::TransformFeedback);
+            } else {
+                LOG_ERROR(Render_Vulkan, "Shader requires transform feedbacks but these are not "
+                                         "supported on this device");
+            }
+        }
+
         if (ir.UsesLayer() || ir.UsesViewportIndex()) {
             if (ir.UsesViewportIndex()) {
                 AddCapability(spv::Capability::MultiViewport);
@@ -295,7 +318,7 @@ public:
             }
         }
 
-        if (device.IsShaderStorageImageReadWithoutFormatSupported()) {
+        if (device.IsFormatlessImageLoadSupported()) {
             AddCapability(spv::Capability::StorageImageReadWithoutFormat);
         }
 
@@ -317,25 +340,29 @@ public:
             AddExecutionMode(main, spv::ExecutionMode::OutputVertices,
                              header.common2.threads_per_input_primitive);
             break;
-        case ShaderType::TesselationEval:
+        case ShaderType::TesselationEval: {
+            const auto& info = registry.GetGraphicsInfo();
             AddCapability(spv::Capability::Tessellation);
             AddEntryPoint(spv::ExecutionModel::TessellationEvaluation, main, "main", interfaces);
-            AddExecutionMode(main, GetExecutionMode(specialization.tessellation.primitive));
-            AddExecutionMode(main, GetExecutionMode(specialization.tessellation.spacing));
-            AddExecutionMode(main, specialization.tessellation.clockwise
+            AddExecutionMode(main, GetExecutionMode(info.tessellation_primitive));
+            AddExecutionMode(main, GetExecutionMode(info.tessellation_spacing));
+            AddExecutionMode(main, info.tessellation_clockwise
                                        ? spv::ExecutionMode::VertexOrderCw
                                        : spv::ExecutionMode::VertexOrderCcw);
             break;
-        case ShaderType::Geometry:
+        }
+        case ShaderType::Geometry: {
+            const auto& info = registry.GetGraphicsInfo();
             AddCapability(spv::Capability::Geometry);
             AddEntryPoint(spv::ExecutionModel::Geometry, main, "main", interfaces);
-            AddExecutionMode(main, GetExecutionMode(specialization.primitive_topology));
+            AddExecutionMode(main, GetExecutionMode(info.primitive_topology));
             AddExecutionMode(main, GetExecutionMode(header.common3.output_topology));
             AddExecutionMode(main, spv::ExecutionMode::OutputVertices,
                              header.common4.max_output_vertices);
             // TODO(Rodrigo): Where can we get this info from?
             AddExecutionMode(main, spv::ExecutionMode::Invocations, 1U);
             break;
+        }
         case ShaderType::Fragment:
             AddEntryPoint(spv::ExecutionModel::Fragment, main, "main", interfaces);
             AddExecutionMode(main, spv::ExecutionMode::OriginUpperLeft);
@@ -544,7 +571,8 @@ private:
         if (stage != ShaderType::Geometry) {
             return;
         }
-        const u32 num_input = GetNumPrimitiveTopologyVertices(specialization.primitive_topology);
+        const auto& info = registry.GetGraphicsInfo();
+        const u32 num_input = GetNumPrimitiveTopologyVertices(info.primitive_topology);
         DeclareInputVertexArray(num_input);
         DeclareOutputVertex();
     }
@@ -741,12 +769,34 @@ private:
     }
 
     void DeclareOutputAttributes() {
+        if (stage == ShaderType::Compute || stage == ShaderType::Fragment) {
+            return;
+        }
+
+        UNIMPLEMENTED_IF(registry.GetGraphicsInfo().tfb_enabled && stage != ShaderType::Vertex);
         for (const auto index : ir.GetOutputAttributes()) {
             if (!IsGenericAttribute(index)) {
                 continue;
             }
-            const u32 location = GetGenericAttributeLocation(index);
-            Id type = t_float4;
+            DeclareOutputAttribute(index);
+        }
+    }
+
+    void DeclareOutputAttribute(Attribute::Index index) {
+        static constexpr std::string_view swizzle = "xyzw";
+
+        const u32 location = GetGenericAttributeLocation(index);
+        u8 element = 0;
+        while (element < 4) {
+            const std::size_t remainder = 4 - element;
+
+            std::size_t num_components = remainder;
+            const std::optional tfb = GetTransformFeedbackInfo(index, element);
+            if (tfb) {
+                num_components = tfb->components;
+            }
+
+            Id type = GetTypeVectorDefinitionLut(Type::Float).at(num_components - 1);
             Id varying_default = v_varying_default;
             if (IsOutputAttributeArray()) {
                 const u32 num = GetNumOutputVertices();
@@ -759,15 +809,47 @@ private:
             }
             type = TypePointer(spv::StorageClass::Output, type);
 
+            std::string name = fmt::format("out_attr{}", location);
+            if (num_components < 4 || element > 0) {
+                name = fmt::format("{}_{}", name, swizzle.substr(element, num_components));
+            }
+
             const Id id = OpVariable(type, spv::StorageClass::Output, varying_default);
-            Name(AddGlobalVariable(id), fmt::format("out_attr{}", location));
-            output_attributes.emplace(index, id);
+            Name(AddGlobalVariable(id), name);
+
+            GenericVaryingDescription description;
+            description.id = id;
+            description.first_element = element;
+            description.is_scalar = num_components == 1;
+            for (u32 i = 0; i < num_components; ++i) {
+                const u8 offset = static_cast<u8>(static_cast<u32>(index) * 4 + element + i);
+                output_attributes.emplace(offset, description);
+            }
             interfaces.push_back(id);
 
             Decorate(id, spv::Decoration::Location, location);
+            if (element > 0) {
+                Decorate(id, spv::Decoration::Component, static_cast<u32>(element));
+            }
+            if (tfb && device.IsExtTransformFeedbackSupported()) {
+                Decorate(id, spv::Decoration::XfbBuffer, static_cast<u32>(tfb->buffer));
+                Decorate(id, spv::Decoration::XfbStride, static_cast<u32>(tfb->stride));
+                Decorate(id, spv::Decoration::Offset, static_cast<u32>(tfb->offset));
+            }
+
+            element = static_cast<u8>(static_cast<std::size_t>(element) + num_components);
         }
     }
 
+    std::optional<VaryingTFB> GetTransformFeedbackInfo(Attribute::Index index, u8 element = 0) {
+        const u8 location = static_cast<u8>(static_cast<u32>(index) * 4 + element);
+        const auto it = transform_feedback.find(location);
+        if (it == transform_feedback.end()) {
+            return {};
+        }
+        return it->second;
+    }
+
     u32 DeclareConstantBuffers(u32 binding) {
         for (const auto& [index, size] : ir.GetConstantBuffers()) {
             const Id type = device.IsKhrUniformBufferStandardLayoutSupported() ? t_cbuf_scalar_ubo
@@ -833,16 +915,20 @@ private:
             constexpr int sampled = 1;
             constexpr auto format = spv::ImageFormat::Unknown;
             const Id image_type = TypeImage(t_float, dim, depth, arrayed, ms, sampled, format);
-            const Id sampled_image_type = TypeSampledImage(image_type);
-            const Id pointer_type =
-                TypePointer(spv::StorageClass::UniformConstant, sampled_image_type);
+            const Id sampler_type = TypeSampledImage(image_type);
+            const Id sampler_pointer_type =
+                TypePointer(spv::StorageClass::UniformConstant, sampler_type);
+            const Id type = sampler.IsIndexed()
+                                ? TypeArray(sampler_type, Constant(t_uint, sampler.Size()))
+                                : sampler_type;
+            const Id pointer_type = TypePointer(spv::StorageClass::UniformConstant, type);
             const Id id = OpVariable(pointer_type, spv::StorageClass::UniformConstant);
             AddGlobalVariable(Name(id, fmt::format("sampler_{}", sampler.GetIndex())));
             Decorate(id, spv::Decoration::Binding, binding++);
             Decorate(id, spv::Decoration::DescriptorSet, DESCRIPTOR_SET);
 
-            sampled_images.emplace(sampler.GetIndex(),
-                                   SampledImage{image_type, sampled_image_type, id});
+            sampled_images.emplace(sampler.GetIndex(), SampledImage{image_type, sampler_type,
+                                                                    sampler_pointer_type, id});
         }
         return binding;
     }
@@ -893,7 +979,7 @@ private:
     u32 GetNumInputVertices() const {
         switch (stage) {
         case ShaderType::Geometry:
-            return GetNumPrimitiveTopologyVertices(specialization.primitive_topology);
+            return GetNumPrimitiveTopologyVertices(registry.GetGraphicsInfo().primitive_topology);
         case ShaderType::TesselationControl:
         case ShaderType::TesselationEval:
             return NumInputPatches;
@@ -1341,8 +1427,14 @@ private:
                 }
                 default:
                     if (IsGenericAttribute(attribute)) {
-                        const Id composite = output_attributes.at(attribute);
-                        return {ArrayPass(t_out_float, composite, {element}), Type::Float};
+                        const u8 offset = static_cast<u8>(static_cast<u8>(attribute) * 4 + element);
+                        const GenericVaryingDescription description = output_attributes.at(offset);
+                        const Id composite = description.id;
+                        std::vector<u32> indices;
+                        if (!description.is_scalar) {
+                            indices.push_back(element - description.first_element);
+                        }
+                        return {ArrayPass(t_out_float, composite, indices), Type::Float};
                     }
                     UNIMPLEMENTED_MSG("Unhandled output attribute: {}",
                                       static_cast<u32>(attribute));
@@ -1525,7 +1617,12 @@ private:
         ASSERT(!meta.sampler.IsBuffer());
 
         const auto& entry = sampled_images.at(meta.sampler.GetIndex());
-        return OpLoad(entry.sampled_image_type, entry.sampler);
+        Id sampler = entry.variable;
+        if (meta.sampler.IsIndexed()) {
+            const Id index = AsInt(Visit(meta.index));
+            sampler = OpAccessChain(entry.sampler_pointer_type, sampler, index);
+        }
+        return OpLoad(entry.sampler_type, sampler);
     }
 
     Id GetTextureImage(Operation operation) {
@@ -1783,7 +1880,7 @@ private:
     }
 
     Expression ImageLoad(Operation operation) {
-        if (!device.IsShaderStorageImageReadWithoutFormatSupported()) {
+        if (!device.IsFormatlessImageLoadSupported()) {
             return {v_float_zero, Type::Float};
         }
 
@@ -2211,16 +2308,14 @@ private:
         switch (specialization.attribute_types.at(location)) {
         case Maxwell::VertexAttribute::Type::SignedNorm:
         case Maxwell::VertexAttribute::Type::UnsignedNorm:
+        case Maxwell::VertexAttribute::Type::UnsignedScaled:
+        case Maxwell::VertexAttribute::Type::SignedScaled:
         case Maxwell::VertexAttribute::Type::Float:
             return {Type::Float, t_in_float, t_in_float4};
         case Maxwell::VertexAttribute::Type::SignedInt:
             return {Type::Int, t_in_int, t_in_int4};
         case Maxwell::VertexAttribute::Type::UnsignedInt:
             return {Type::Uint, t_in_uint, t_in_uint4};
-        case Maxwell::VertexAttribute::Type::UnsignedScaled:
-        case Maxwell::VertexAttribute::Type::SignedScaled:
-            UNIMPLEMENTED();
-            return {Type::Float, t_in_float, t_in_float4};
         default:
             UNREACHABLE();
             return {Type::Float, t_in_float, t_in_float4};
@@ -2250,11 +2345,11 @@ private:
     std::array<Id, 4> GetTypeVectorDefinitionLut(Type type) const {
         switch (type) {
         case Type::Float:
-            return {nullptr, t_float2, t_float3, t_float4};
+            return {t_float, t_float2, t_float3, t_float4};
         case Type::Int:
-            return {nullptr, t_int2, t_int3, t_int4};
+            return {t_int, t_int2, t_int3, t_int4};
         case Type::Uint:
-            return {nullptr, t_uint2, t_uint3, t_uint4};
+            return {t_uint, t_uint2, t_uint3, t_uint4};
         default:
             UNIMPLEMENTED();
             return {};
@@ -2487,7 +2582,9 @@ private:
     const ShaderIR& ir;
     const ShaderType stage;
     const Tegra::Shader::Header header;
+    const Registry& registry;
     const Specialization& specialization;
+    std::unordered_map<u8, VaryingTFB> transform_feedback;
 
     const Id t_void = Name(TypeVoid(), "void");
 
@@ -2576,7 +2673,7 @@ private:
     Id shared_memory{};
     std::array<Id, INTERNAL_FLAGS_COUNT> internal_flags{};
     std::map<Attribute::Index, Id> input_attributes;
-    std::map<Attribute::Index, Id> output_attributes;
+    std::unordered_map<u8, GenericVaryingDescription> output_attributes;
     std::map<u32, Id> constant_buffers;
     std::map<GlobalMemoryBase, Id> global_buffers;
     std::map<u32, TexelBuffer> texel_buffers;
@@ -2862,8 +2959,9 @@ ShaderEntries GenerateShaderEntries(const VideoCommon::Shader::ShaderIR& ir) {
 }
 
 std::vector<u32> Decompile(const VKDevice& device, const VideoCommon::Shader::ShaderIR& ir,
-                           ShaderType stage, const Specialization& specialization) {
-    return SPIRVDecompiler(device, ir, stage, specialization).Assemble();
+                           ShaderType stage, const VideoCommon::Shader::Registry& registry,
+                           const Specialization& specialization) {
+    return SPIRVDecompiler(device, ir, stage, registry, specialization).Assemble();
 }
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.h b/src/video_core/renderer_vulkan/vk_shader_decompiler.h
index f5dc14d9e..ffea4709e 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.h
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.h
@@ -15,6 +15,7 @@
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/shader_type.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace Vulkan {
@@ -91,17 +92,9 @@ struct Specialization final {
     u32 shared_memory_size{};
 
     // Graphics specific
-    Maxwell::PrimitiveTopology primitive_topology{};
     std::optional<float> point_size{};
     std::array<Maxwell::VertexAttribute::Type, Maxwell::NumVertexAttributes> attribute_types{};
     bool ndc_minus_one_to_one{};
-
-    // Tessellation specific
-    struct {
-        Maxwell::TessellationPrimitive primitive{};
-        Maxwell::TessellationSpacing spacing{};
-        bool clockwise{};
-    } tessellation;
 };
 // Old gcc versions don't consider this trivially copyable.
 // static_assert(std::is_trivially_copyable_v<Specialization>);
@@ -114,6 +107,8 @@ struct SPIRVShader {
 ShaderEntries GenerateShaderEntries(const VideoCommon::Shader::ShaderIR& ir);
 
 std::vector<u32> Decompile(const VKDevice& device, const VideoCommon::Shader::ShaderIR& ir,
-                           Tegra::Engines::ShaderType stage, const Specialization& specialization);
+                           Tegra::Engines::ShaderType stage,
+                           const VideoCommon::Shader::Registry& registry,
+                           const Specialization& specialization);
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
index 171d78afc..374959f82 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
@@ -73,7 +73,8 @@ VKBuffer* VKStagingBufferPool::TryGetReservedBuffer(std::size_t size, bool host_
 VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_visible) {
     const auto usage =
         vk::BufferUsageFlagBits::eTransferSrc | vk::BufferUsageFlagBits::eTransferDst |
-        vk::BufferUsageFlagBits::eStorageBuffer | vk::BufferUsageFlagBits::eIndexBuffer;
+        vk::BufferUsageFlagBits::eUniformBuffer | vk::BufferUsageFlagBits::eStorageBuffer |
+        vk::BufferUsageFlagBits::eIndexBuffer;
     const u32 log2 = Common::Log2Ceil64(size);
     const vk::BufferCreateInfo buffer_ci({}, 1ULL << log2, usage, vk::SharingMode::eExclusive, 0,
                                          nullptr);
@@ -99,7 +100,6 @@ void VKStagingBufferPool::ReleaseCache(bool host_visible) {
 }
 
 u64 VKStagingBufferPool::ReleaseLevel(StagingBuffersCache& cache, std::size_t log2) {
-    static constexpr u64 epochs_to_destroy = 180;
     static constexpr std::size_t deletions_per_tick = 16;
 
     auto& staging = cache[log2];
@@ -107,6 +107,7 @@ u64 VKStagingBufferPool::ReleaseLevel(StagingBuffersCache& cache, std::size_t lo
     const std::size_t old_size = entries.size();
 
     const auto is_deleteable = [this](const auto& entry) {
+        static constexpr u64 epochs_to_destroy = 180;
         return entry.last_epoch + epochs_to_destroy < epoch && !entry.watch.IsUsed();
     };
     const std::size_t begin_offset = staging.delete_index;
diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.cpp b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
new file mode 100644
index 000000000..94a89e388
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_state_tracker.cpp
@@ -0,0 +1,99 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstddef>
+#include <iterator>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/dirty_flags.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/renderer_vulkan/vk_state_tracker.h"
+
+#define OFF(field_name) MAXWELL3D_REG_INDEX(field_name)
+#define NUM(field_name) (sizeof(Maxwell3D::Regs::field_name) / sizeof(u32))
+
+namespace Vulkan {
+
+namespace {
+
+using namespace Dirty;
+using namespace VideoCommon::Dirty;
+using Tegra::Engines::Maxwell3D;
+using Regs = Maxwell3D::Regs;
+using Tables = Maxwell3D::DirtyState::Tables;
+using Table = Maxwell3D::DirtyState::Table;
+using Flags = Maxwell3D::DirtyState::Flags;
+
+Flags MakeInvalidationFlags() {
+    Flags flags{};
+    flags[Viewports] = true;
+    flags[Scissors] = true;
+    flags[DepthBias] = true;
+    flags[BlendConstants] = true;
+    flags[DepthBounds] = true;
+    flags[StencilProperties] = true;
+    return flags;
+}
+
+void SetupDirtyViewports(Tables& tables) {
+    FillBlock(tables[0], OFF(viewport_transform), NUM(viewport_transform), Viewports);
+    FillBlock(tables[0], OFF(viewports), NUM(viewports), Viewports);
+    tables[0][OFF(viewport_transform_enabled)] = Viewports;
+}
+
+void SetupDirtyScissors(Tables& tables) {
+    FillBlock(tables[0], OFF(scissor_test), NUM(scissor_test), Scissors);
+}
+
+void SetupDirtyDepthBias(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(polygon_offset_units)] = DepthBias;
+    table[OFF(polygon_offset_clamp)] = DepthBias;
+    table[OFF(polygon_offset_factor)] = DepthBias;
+}
+
+void SetupDirtyBlendConstants(Tables& tables) {
+    FillBlock(tables[0], OFF(blend_color), NUM(blend_color), BlendConstants);
+}
+
+void SetupDirtyDepthBounds(Tables& tables) {
+    FillBlock(tables[0], OFF(depth_bounds), NUM(depth_bounds), DepthBounds);
+}
+
+void SetupDirtyStencilProperties(Tables& tables) {
+    auto& table = tables[0];
+    table[OFF(stencil_two_side_enable)] = StencilProperties;
+    table[OFF(stencil_front_func_ref)] = StencilProperties;
+    table[OFF(stencil_front_mask)] = StencilProperties;
+    table[OFF(stencil_front_func_mask)] = StencilProperties;
+    table[OFF(stencil_back_func_ref)] = StencilProperties;
+    table[OFF(stencil_back_mask)] = StencilProperties;
+    table[OFF(stencil_back_func_mask)] = StencilProperties;
+}
+
+} // Anonymous namespace
+
+StateTracker::StateTracker(Core::System& system)
+    : system{system}, invalidation_flags{MakeInvalidationFlags()} {}
+
+void StateTracker::Initialize() {
+    auto& dirty = system.GPU().Maxwell3D().dirty;
+    auto& tables = dirty.tables;
+    SetupDirtyRenderTargets(tables);
+    SetupDirtyViewports(tables);
+    SetupDirtyScissors(tables);
+    SetupDirtyDepthBias(tables);
+    SetupDirtyBlendConstants(tables);
+    SetupDirtyDepthBounds(tables);
+    SetupDirtyStencilProperties(tables);
+}
+
+void StateTracker::InvalidateCommandBufferState() {
+    system.GPU().Maxwell3D().dirty.flags |= invalidation_flags;
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_state_tracker.h b/src/video_core/renderer_vulkan/vk_state_tracker.h
new file mode 100644
index 000000000..03bc415b2
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_state_tracker.h
@@ -0,0 +1,79 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <limits>
+
+#include "common/common_types.h"
+#include "core/core.h"
+#include "video_core/dirty_flags.h"
+#include "video_core/engines/maxwell_3d.h"
+
+namespace Vulkan {
+
+namespace Dirty {
+
+enum : u8 {
+    First = VideoCommon::Dirty::LastCommonEntry,
+
+    Viewports,
+    Scissors,
+    DepthBias,
+    BlendConstants,
+    DepthBounds,
+    StencilProperties,
+
+    Last
+};
+static_assert(Last <= std::numeric_limits<u8>::max());
+
+} // namespace Dirty
+
+class StateTracker {
+public:
+    explicit StateTracker(Core::System& system);
+
+    void Initialize();
+
+    void InvalidateCommandBufferState();
+
+    bool TouchViewports() {
+        return Exchange(Dirty::Viewports, false);
+    }
+
+    bool TouchScissors() {
+        return Exchange(Dirty::Scissors, false);
+    }
+
+    bool TouchDepthBias() {
+        return Exchange(Dirty::DepthBias, false);
+    }
+
+    bool TouchBlendConstants() {
+        return Exchange(Dirty::BlendConstants, false);
+    }
+
+    bool TouchDepthBounds() {
+        return Exchange(Dirty::DepthBounds, false);
+    }
+
+    bool TouchStencilProperties() {
+        return Exchange(Dirty::StencilProperties, false);
+    }
+
+private:
+    bool Exchange(std::size_t id, bool new_value) const noexcept {
+        auto& flags = system.GPU().Maxwell3D().dirty.flags;
+        const bool is_dirty = flags[id];
+        flags[id] = new_value;
+        return is_dirty;
+    }
+
+    Core::System& system;
+    Tegra::Engines::Maxwell3D::DirtyState::Flags invalidation_flags;
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_swapchain.cpp b/src/video_core/renderer_vulkan/vk_swapchain.cpp
index f47b691a8..9e73fa9cd 100644
--- a/src/video_core/renderer_vulkan/vk_swapchain.cpp
+++ b/src/video_core/renderer_vulkan/vk_swapchain.cpp
@@ -141,11 +141,6 @@ void VKSwapchain::CreateSwapchain(const vk::SurfaceCapabilitiesKHR& capabilities
 
     const vk::SurfaceFormatKHR surface_format{ChooseSwapSurfaceFormat(formats, srgb)};
     const vk::PresentModeKHR present_mode{ChooseSwapPresentMode(present_modes)};
-    extent = ChooseSwapExtent(capabilities, width, height);
-
-    current_width = extent.width;
-    current_height = extent.height;
-    current_srgb = srgb;
 
     u32 requested_image_count{capabilities.minImageCount + 1};
     if (capabilities.maxImageCount > 0 && requested_image_count > capabilities.maxImageCount) {
@@ -153,10 +148,9 @@ void VKSwapchain::CreateSwapchain(const vk::SurfaceCapabilitiesKHR& capabilities
     }
 
     vk::SwapchainCreateInfoKHR swapchain_ci(
-        {}, surface, requested_image_count, surface_format.format, surface_format.colorSpace,
-        extent, 1, vk::ImageUsageFlagBits::eColorAttachment, {}, {}, {},
-        capabilities.currentTransform, vk::CompositeAlphaFlagBitsKHR::eOpaque, present_mode, false,
-        {});
+        {}, surface, requested_image_count, surface_format.format, surface_format.colorSpace, {}, 1,
+        vk::ImageUsageFlagBits::eColorAttachment, {}, {}, {}, capabilities.currentTransform,
+        vk::CompositeAlphaFlagBitsKHR::eOpaque, present_mode, false, {});
 
     const u32 graphics_family{device.GetGraphicsFamily()};
     const u32 present_family{device.GetPresentFamily()};
@@ -169,9 +163,18 @@ void VKSwapchain::CreateSwapchain(const vk::SurfaceCapabilitiesKHR& capabilities
         swapchain_ci.imageSharingMode = vk::SharingMode::eExclusive;
     }
 
+    // Request the size again to reduce the possibility of a TOCTOU race condition.
+    const auto updated_capabilities = physical_device.getSurfaceCapabilitiesKHR(surface, dld);
+    swapchain_ci.imageExtent = ChooseSwapExtent(updated_capabilities, width, height);
+    // Don't add code within this and the swapchain creation.
     const auto dev{device.GetLogical()};
     swapchain = dev.createSwapchainKHRUnique(swapchain_ci, nullptr, dld);
 
+    extent = swapchain_ci.imageExtent;
+    current_width = extent.width;
+    current_height = extent.height;
+    current_srgb = srgb;
+
     images = dev.getSwapchainImagesKHR(*swapchain, dld);
     image_count = static_cast<u32>(images.size());
     image_format = surface_format.format;
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.cpp b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
index 51b0d38a6..26175921b 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.cpp
@@ -22,6 +22,7 @@
 #include "video_core/renderer_vulkan/vk_device.h"
 #include "video_core/renderer_vulkan/vk_memory_manager.h"
 #include "video_core/renderer_vulkan/vk_rasterizer.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
 #include "video_core/renderer_vulkan/vk_texture_cache.h"
 #include "video_core/surface.h"
@@ -51,6 +52,9 @@ vk::ImageType SurfaceTargetToImage(SurfaceTarget target) {
         return vk::ImageType::e2D;
     case SurfaceTarget::Texture3D:
         return vk::ImageType::e3D;
+    case SurfaceTarget::TextureBuffer:
+        UNREACHABLE();
+        return {};
     }
     UNREACHABLE_MSG("Unknown texture target={}", static_cast<u32>(target));
     return {};
@@ -272,7 +276,6 @@ void CachedSurface::UploadImage(const std::vector<u8>& staging_buffer) {
 
     for (u32 level = 0; level < params.num_levels; ++level) {
         vk::BufferImageCopy copy = GetBufferImageCopy(level);
-        const auto& dld = device.GetDispatchLoader();
         if (image->GetAspectMask() ==
             (vk::ImageAspectFlagBits::eDepth | vk::ImageAspectFlagBits::eStencil)) {
             vk::BufferImageCopy depth = copy;
@@ -421,7 +424,6 @@ void VKTextureCache::ImageCopy(Surface& src_surface, Surface& dst_surface,
         dst_base_layer, num_layers, copy_params.dest_level, 1, vk::PipelineStageFlagBits::eTransfer,
         vk::AccessFlagBits::eTransferWrite, vk::ImageLayout::eTransferDstOptimal);
 
-    const auto& dld{device.GetDispatchLoader()};
     const vk::ImageSubresourceLayers src_subresource(
         src_surface->GetAspectMask(), copy_params.source_level, copy_params.source_z, num_layers);
     const vk::ImageSubresourceLayers dst_subresource(
@@ -457,7 +459,6 @@ void VKTextureCache::ImageBlit(View& src_view, View& dst_view,
                              dst_view->GetImageSubresourceLayers(), {dst_top_left, dst_bot_right});
     const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear;
 
-    const auto& dld{device.GetDispatchLoader()};
     scheduler.Record([src_image = src_view->GetImage(), dst_image = dst_view->GetImage(), blit,
                       is_linear](auto cmdbuf, auto& dld) {
         cmdbuf.blitImage(src_image, vk::ImageLayout::eTransferSrcOptimal, dst_image,
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h
index d3edbe80c..22e3d34de 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.h
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.h
@@ -151,6 +151,10 @@ public:
         return params.GetMipHeight(base_level);
     }
 
+    u32 GetNumLayers() const {
+        return num_layers;
+    }
+
     bool IsBufferView() const {
         return buffer_view;
     }
diff --git a/src/video_core/shader/const_buffer_locker.cpp b/src/video_core/shader/const_buffer_locker.cpp
deleted file mode 100644
index 0638be8cb..000000000
--- a/src/video_core/shader/const_buffer_locker.cpp
+++ /dev/null
@@ -1,126 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <algorithm>
-#include <tuple>
-
-#include "common/common_types.h"
-#include "video_core/engines/maxwell_3d.h"
-#include "video_core/engines/shader_type.h"
-#include "video_core/shader/const_buffer_locker.h"
-
-namespace VideoCommon::Shader {
-
-using Tegra::Engines::SamplerDescriptor;
-
-ConstBufferLocker::ConstBufferLocker(Tegra::Engines::ShaderType shader_stage)
-    : stage{shader_stage} {}
-
-ConstBufferLocker::ConstBufferLocker(Tegra::Engines::ShaderType shader_stage,
-                                     Tegra::Engines::ConstBufferEngineInterface& engine)
-    : stage{shader_stage}, engine{&engine} {}
-
-ConstBufferLocker::~ConstBufferLocker() = default;
-
-std::optional<u32> ConstBufferLocker::ObtainKey(u32 buffer, u32 offset) {
-    const std::pair<u32, u32> key = {buffer, offset};
-    const auto iter = keys.find(key);
-    if (iter != keys.end()) {
-        return iter->second;
-    }
-    if (!engine) {
-        return std::nullopt;
-    }
-    const u32 value = engine->AccessConstBuffer32(stage, buffer, offset);
-    keys.emplace(key, value);
-    return value;
-}
-
-std::optional<SamplerDescriptor> ConstBufferLocker::ObtainBoundSampler(u32 offset) {
-    const u32 key = offset;
-    const auto iter = bound_samplers.find(key);
-    if (iter != bound_samplers.end()) {
-        return iter->second;
-    }
-    if (!engine) {
-        return std::nullopt;
-    }
-    const SamplerDescriptor value = engine->AccessBoundSampler(stage, offset);
-    bound_samplers.emplace(key, value);
-    return value;
-}
-
-std::optional<Tegra::Engines::SamplerDescriptor> ConstBufferLocker::ObtainBindlessSampler(
-    u32 buffer, u32 offset) {
-    const std::pair key = {buffer, offset};
-    const auto iter = bindless_samplers.find(key);
-    if (iter != bindless_samplers.end()) {
-        return iter->second;
-    }
-    if (!engine) {
-        return std::nullopt;
-    }
-    const SamplerDescriptor value = engine->AccessBindlessSampler(stage, buffer, offset);
-    bindless_samplers.emplace(key, value);
-    return value;
-}
-
-std::optional<u32> ConstBufferLocker::ObtainBoundBuffer() {
-    if (bound_buffer_saved) {
-        return bound_buffer;
-    }
-    if (!engine) {
-        return std::nullopt;
-    }
-    bound_buffer_saved = true;
-    bound_buffer = engine->GetBoundBuffer();
-    return bound_buffer;
-}
-
-void ConstBufferLocker::InsertKey(u32 buffer, u32 offset, u32 value) {
-    keys.insert_or_assign({buffer, offset}, value);
-}
-
-void ConstBufferLocker::InsertBoundSampler(u32 offset, SamplerDescriptor sampler) {
-    bound_samplers.insert_or_assign(offset, sampler);
-}
-
-void ConstBufferLocker::InsertBindlessSampler(u32 buffer, u32 offset, SamplerDescriptor sampler) {
-    bindless_samplers.insert_or_assign({buffer, offset}, sampler);
-}
-
-void ConstBufferLocker::SetBoundBuffer(u32 buffer) {
-    bound_buffer_saved = true;
-    bound_buffer = buffer;
-}
-
-bool ConstBufferLocker::IsConsistent() const {
-    if (!engine) {
-        return false;
-    }
-    return std::all_of(keys.begin(), keys.end(),
-                       [this](const auto& pair) {
-                           const auto [cbuf, offset] = pair.first;
-                           const auto value = pair.second;
-                           return value == engine->AccessConstBuffer32(stage, cbuf, offset);
-                       }) &&
-           std::all_of(bound_samplers.begin(), bound_samplers.end(),
-                       [this](const auto& sampler) {
-                           const auto [key, value] = sampler;
-                           return value == engine->AccessBoundSampler(stage, key);
-                       }) &&
-           std::all_of(bindless_samplers.begin(), bindless_samplers.end(),
-                       [this](const auto& sampler) {
-                           const auto [cbuf, offset] = sampler.first;
-                           const auto value = sampler.second;
-                           return value == engine->AccessBindlessSampler(stage, cbuf, offset);
-                       });
-}
-
-bool ConstBufferLocker::HasEqualKeys(const ConstBufferLocker& rhs) const {
-    return std::tie(keys, bound_samplers, bindless_samplers) ==
-           std::tie(rhs.keys, rhs.bound_samplers, rhs.bindless_samplers);
-}
-
-} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/const_buffer_locker.h b/src/video_core/shader/const_buffer_locker.h
deleted file mode 100644
index d3ea11087..000000000
--- a/src/video_core/shader/const_buffer_locker.h
+++ /dev/null
@@ -1,103 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <optional>
-#include <unordered_map>
-#include "common/common_types.h"
-#include "common/hash.h"
-#include "video_core/engines/const_buffer_engine_interface.h"
-#include "video_core/engines/shader_type.h"
-#include "video_core/guest_driver.h"
-
-namespace VideoCommon::Shader {
-
-using KeyMap = std::unordered_map<std::pair<u32, u32>, u32, Common::PairHash>;
-using BoundSamplerMap = std::unordered_map<u32, Tegra::Engines::SamplerDescriptor>;
-using BindlessSamplerMap =
-    std::unordered_map<std::pair<u32, u32>, Tegra::Engines::SamplerDescriptor, Common::PairHash>;
-
-/**
- * The ConstBufferLocker is a class use to interface the 3D and compute engines with the shader
- * compiler. with it, the shader can obtain required data from GPU state and store it for disk
- * shader compilation.
- */
-class ConstBufferLocker {
-public:
-    explicit ConstBufferLocker(Tegra::Engines::ShaderType shader_stage);
-
-    explicit ConstBufferLocker(Tegra::Engines::ShaderType shader_stage,
-                               Tegra::Engines::ConstBufferEngineInterface& engine);
-
-    ~ConstBufferLocker();
-
-    /// Retrieves a key from the locker, if it's registered, it will give the registered value, if
-    /// not it will obtain it from maxwell3d and register it.
-    std::optional<u32> ObtainKey(u32 buffer, u32 offset);
-
-    std::optional<Tegra::Engines::SamplerDescriptor> ObtainBoundSampler(u32 offset);
-
-    std::optional<Tegra::Engines::SamplerDescriptor> ObtainBindlessSampler(u32 buffer, u32 offset);
-
-    std::optional<u32> ObtainBoundBuffer();
-
-    /// Inserts a key.
-    void InsertKey(u32 buffer, u32 offset, u32 value);
-
-    /// Inserts a bound sampler key.
-    void InsertBoundSampler(u32 offset, Tegra::Engines::SamplerDescriptor sampler);
-
-    /// Inserts a bindless sampler key.
-    void InsertBindlessSampler(u32 buffer, u32 offset, Tegra::Engines::SamplerDescriptor sampler);
-
-    /// Set the bound buffer for this locker.
-    void SetBoundBuffer(u32 buffer);
-
-    /// Checks keys and samplers against engine's current const buffers. Returns true if they are
-    /// the same value, false otherwise;
-    bool IsConsistent() const;
-
-    /// Returns true if the keys are equal to the other ones in the locker.
-    bool HasEqualKeys(const ConstBufferLocker& rhs) const;
-
-    /// Gives an getter to the const buffer keys in the database.
-    const KeyMap& GetKeys() const {
-        return keys;
-    }
-
-    /// Gets samplers database.
-    const BoundSamplerMap& GetBoundSamplers() const {
-        return bound_samplers;
-    }
-
-    /// Gets bindless samplers database.
-    const BindlessSamplerMap& GetBindlessSamplers() const {
-        return bindless_samplers;
-    }
-
-    /// Gets bound buffer used on this shader
-    u32 GetBoundBuffer() const {
-        return bound_buffer;
-    }
-
-    /// Obtains access to the guest driver's profile.
-    VideoCore::GuestDriverProfile* AccessGuestDriverProfile() const {
-        if (engine) {
-            return &engine->AccessGuestDriverProfile();
-        }
-        return nullptr;
-    }
-
-private:
-    const Tegra::Engines::ShaderType stage;
-    Tegra::Engines::ConstBufferEngineInterface* engine = nullptr;
-    KeyMap keys;
-    BoundSamplerMap bound_samplers;
-    BindlessSamplerMap bindless_samplers;
-    bool bound_buffer_saved{};
-    u32 bound_buffer{};
-};
-
-} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/control_flow.cpp b/src/video_core/shader/control_flow.cpp
index 0229733b6..2e2711350 100644
--- a/src/video_core/shader/control_flow.cpp
+++ b/src/video_core/shader/control_flow.cpp
@@ -13,6 +13,7 @@
 #include "common/common_types.h"
 #include "video_core/shader/ast.h"
 #include "video_core/shader/control_flow.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace VideoCommon::Shader {
@@ -64,11 +65,11 @@ struct BlockInfo {
 };
 
 struct CFGRebuildState {
-    explicit CFGRebuildState(const ProgramCode& program_code, u32 start, ConstBufferLocker& locker)
-        : program_code{program_code}, locker{locker}, start{start} {}
+    explicit CFGRebuildState(const ProgramCode& program_code, u32 start, Registry& registry)
+        : program_code{program_code}, registry{registry}, start{start} {}
 
     const ProgramCode& program_code;
-    ConstBufferLocker& locker;
+    Registry& registry;
     u32 start{};
     std::vector<BlockInfo> block_info;
     std::list<u32> inspect_queries;
@@ -438,7 +439,7 @@ std::pair<ParseResult, ParseInfo> ParseCode(CFGRebuildState& state, u32 address)
             const s32 pc_target = offset + result.relative_position;
             std::vector<CaseBranch> branches;
             for (u32 i = 0; i < result.entries; i++) {
-                auto key = state.locker.ObtainKey(result.buffer, result.offset + i * 4);
+                auto key = state.registry.ObtainKey(result.buffer, result.offset + i * 4);
                 if (!key) {
                     return {ParseResult::AbnormalFlow, parse_info};
                 }
@@ -656,14 +657,14 @@ void DecompileShader(CFGRebuildState& state) {
 
 std::unique_ptr<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u32 start_address,
                                                 const CompilerSettings& settings,
-                                                ConstBufferLocker& locker) {
+                                                Registry& registry) {
     auto result_out = std::make_unique<ShaderCharacteristics>();
     if (settings.depth == CompileDepth::BruteForce) {
         result_out->settings.depth = CompileDepth::BruteForce;
         return result_out;
     }
 
-    CFGRebuildState state{program_code, start_address, locker};
+    CFGRebuildState state{program_code, start_address, registry};
     // Inspect Code and generate blocks
     state.labels.clear();
     state.labels.emplace(start_address);
diff --git a/src/video_core/shader/control_flow.h b/src/video_core/shader/control_flow.h
index 5304998b9..62a3510d8 100644
--- a/src/video_core/shader/control_flow.h
+++ b/src/video_core/shader/control_flow.h
@@ -12,6 +12,7 @@
 #include "video_core/engines/shader_bytecode.h"
 #include "video_core/shader/ast.h"
 #include "video_core/shader/compiler_settings.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace VideoCommon::Shader {
@@ -111,6 +112,6 @@ struct ShaderCharacteristics {
 
 std::unique_ptr<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u32 start_address,
                                                 const CompilerSettings& settings,
-                                                ConstBufferLocker& locker);
+                                                Registry& registry);
 
 } // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp
index 6b697ed5d..87ac9ac6c 100644
--- a/src/video_core/shader/decode.cpp
+++ b/src/video_core/shader/decode.cpp
@@ -34,13 +34,9 @@ constexpr bool IsSchedInstruction(u32 offset, u32 main_offset) {
     return (absolute_offset % SchedPeriod) == 0;
 }
 
-void DeduceTextureHandlerSize(VideoCore::GuestDriverProfile* gpu_driver,
+void DeduceTextureHandlerSize(VideoCore::GuestDriverProfile& gpu_driver,
                               const std::list<Sampler>& used_samplers) {
-    if (gpu_driver == nullptr) {
-        LOG_CRITICAL(HW_GPU, "GPU driver profile has not been created yet");
-        return;
-    }
-    if (gpu_driver->TextureHandlerSizeKnown() || used_samplers.size() <= 1) {
+    if (gpu_driver.IsTextureHandlerSizeKnown() || used_samplers.size() <= 1) {
         return;
     }
     u32 count{};
@@ -53,17 +49,13 @@ void DeduceTextureHandlerSize(VideoCore::GuestDriverProfile* gpu_driver,
         bound_offsets.emplace_back(sampler.GetOffset());
     }
     if (count > 1) {
-        gpu_driver->DeduceTextureHandlerSize(std::move(bound_offsets));
+        gpu_driver.DeduceTextureHandlerSize(std::move(bound_offsets));
     }
 }
 
 std::optional<u32> TryDeduceSamplerSize(const Sampler& sampler_to_deduce,
-                                        VideoCore::GuestDriverProfile* gpu_driver,
+                                        VideoCore::GuestDriverProfile& gpu_driver,
                                         const std::list<Sampler>& used_samplers) {
-    if (gpu_driver == nullptr) {
-        LOG_CRITICAL(HW_GPU, "GPU Driver profile has not been created yet");
-        return std::nullopt;
-    }
     const u32 base_offset = sampler_to_deduce.GetOffset();
     u32 max_offset{std::numeric_limits<u32>::max()};
     for (const auto& sampler : used_samplers) {
@@ -77,7 +69,7 @@ std::optional<u32> TryDeduceSamplerSize(const Sampler& sampler_to_deduce,
     if (max_offset == std::numeric_limits<u32>::max()) {
         return std::nullopt;
     }
-    return ((max_offset - base_offset) * 4) / gpu_driver->GetTextureHandlerSize();
+    return ((max_offset - base_offset) * 4) / gpu_driver.GetTextureHandlerSize();
 }
 
 } // Anonymous namespace
@@ -149,7 +141,7 @@ void ShaderIR::Decode() {
     std::memcpy(&header, program_code.data(), sizeof(Tegra::Shader::Header));
 
     decompiled = false;
-    auto info = ScanFlow(program_code, main_offset, settings, locker);
+    auto info = ScanFlow(program_code, main_offset, settings, registry);
     auto& shader_info = *info;
     coverage_begin = shader_info.start;
     coverage_end = shader_info.end;
@@ -364,7 +356,7 @@ u32 ShaderIR::DecodeInstr(NodeBlock& bb, u32 pc) {
 
 void ShaderIR::PostDecode() {
     // Deduce texture handler size if needed
-    auto gpu_driver = locker.AccessGuestDriverProfile();
+    auto gpu_driver = registry.AccessGuestDriverProfile();
     DeduceTextureHandlerSize(gpu_driver, used_samplers);
     // Deduce Indexed Samplers
     if (!uses_indexed_samplers) {
diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp
index 90240c765..478394682 100644
--- a/src/video_core/shader/decode/arithmetic.cpp
+++ b/src/video_core/shader/decode/arithmetic.cpp
@@ -53,29 +53,24 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) {
 
         op_b = GetOperandAbsNegFloat(op_b, false, instr.fmul.negate_b);
 
-        // TODO(Rodrigo): Should precise be used when there's a postfactor?
-        Node value = Operation(OperationCode::FMul, PRECISE, op_a, op_b);
+        static constexpr std::array FmulPostFactor = {
+            1.000f, // None
+            0.500f, // Divide 2
+            0.250f, // Divide 4
+            0.125f, // Divide 8
+            8.000f, // Mul 8
+            4.000f, // Mul 4
+            2.000f, // Mul 2
+        };
 
         if (instr.fmul.postfactor != 0) {
-            auto postfactor = static_cast<s32>(instr.fmul.postfactor);
-
-            // Postfactor encoded as 3-bit 1's complement in instruction, interpreted with below
-            // logic.
-            if (postfactor >= 4) {
-                postfactor = 7 - postfactor;
-            } else {
-                postfactor = 0 - postfactor;
-            }
-
-            if (postfactor > 0) {
-                value = Operation(OperationCode::FMul, NO_PRECISE, value,
-                                  Immediate(static_cast<f32>(1 << postfactor)));
-            } else {
-                value = Operation(OperationCode::FDiv, NO_PRECISE, value,
-                                  Immediate(static_cast<f32>(1 << -postfactor)));
-            }
+            op_a = Operation(OperationCode::FMul, NO_PRECISE, op_a,
+                             Immediate(FmulPostFactor[instr.fmul.postfactor]));
         }
 
+        // TODO(Rodrigo): Should precise be used when there's a postfactor?
+        Node value = Operation(OperationCode::FMul, PRECISE, op_a, op_b);
+
         value = GetSaturatedFloat(value, instr.alu.saturate_d);
 
         SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index 21366869d..2fe787d6f 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -293,44 +293,66 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
 
 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);
-    const Node two = Immediate(2);
-
-    Node value;
-    for (u32 i = 0; i < lop_iterations; ++i) {
-        const Node shift_amount = Immediate(i);
-
-        const Node a = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_c, shift_amount);
-        const Node pack_0 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, one);
-
-        const Node b = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_b, shift_amount);
-        const Node c = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, b, one);
-        const Node pack_1 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, c, one);
-
-        const Node d = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_a, shift_amount);
-        const Node e = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, d, one);
-        const Node pack_2 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, e, two);
-
-        const Node pack_01 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_0, pack_1);
-        const Node pack_012 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_01, pack_2);
-
-        const Node shifted_bit =
-            Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, imm_lut, pack_012);
-        const Node bit = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, shifted_bit, one);
-
-        const Node right =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, bit, shift_amount);
-
-        if (i > 0) {
-            value = Operation(OperationCode::IBitwiseOr, NO_PRECISE, value, right);
-        } else {
-            value = right;
+    const Node lop3_fast = [&](const Node na, const Node nb, const Node nc, const Node ttbl) {
+        Node value = Immediate(0);
+        const ImmediateNode imm = std::get<ImmediateNode>(*ttbl);
+        if (imm.GetValue() & 0x01) {
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, b);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
         }
-    }
+        if (imm.GetValue() & 0x02) {
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, b);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x04) {
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x08) {
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x10) {
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, b);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x20) {
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, b);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x40) {
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x80) {
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        return value;
+    }(op_a, op_b, op_c, imm_lut);
 
-    SetInternalFlagsFromInteger(bb, value, sets_cc);
-    SetRegister(bb, dest, value);
+    SetInternalFlagsFromInteger(bb, lop3_fast, sets_cc);
+    SetRegister(bb, dest, lop3_fast);
 }
 
 } // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/bfe.cpp b/src/video_core/shader/decode/bfe.cpp
index e02bcd097..8e3b46e8e 100644
--- a/src/video_core/shader/decode/bfe.cpp
+++ b/src/video_core/shader/decode/bfe.cpp
@@ -17,33 +17,60 @@ u32 ShaderIR::DecodeBfe(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
-    UNIMPLEMENTED_IF(instr.bfe.negate_b);
-
     Node op_a = GetRegister(instr.gpr8);
-    op_a = GetOperandAbsNegInteger(op_a, false, instr.bfe.negate_a, false);
-
-    switch (opcode->get().GetId()) {
-    case OpCode::Id::BFE_IMM: {
-        UNIMPLEMENTED_IF_MSG(instr.generates_cc,
-                             "Condition codes generation in BFE is not implemented");
+    Node op_b = [&] {
+        switch (opcode->get().GetId()) {
+        case OpCode::Id::BFE_R:
+            return GetRegister(instr.gpr20);
+        case OpCode::Id::BFE_C:
+            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
+        case OpCode::Id::BFE_IMM:
+            return Immediate(instr.alu.GetSignedImm20_20());
+        default:
+            UNREACHABLE();
+            return Immediate(0);
+        }
+    }();
 
-        const Node inner_shift_imm = Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue()));
-        const Node outer_shift_imm =
-            Immediate(static_cast<u32>(instr.bfe.GetLeftShiftValue() + instr.bfe.shift_position));
+    UNIMPLEMENTED_IF_MSG(instr.bfe.rd_cc, "Condition codes in BFE is not implemented");
 
-        const Node inner_shift =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, op_a, inner_shift_imm);
-        const Node outer_shift =
-            Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, inner_shift, outer_shift_imm);
+    const bool is_signed = instr.bfe.is_signed;
 
-        SetInternalFlagsFromInteger(bb, outer_shift, instr.generates_cc);
-        SetRegister(bb, instr.gpr0, outer_shift);
-        break;
-    }
-    default:
-        UNIMPLEMENTED_MSG("Unhandled BFE instruction: {}", opcode->get().GetName());
+    // using reverse parallel method in
+    // https://graphics.stanford.edu/~seander/bithacks.html#ReverseParallel
+    // note for later if possible to implement faster method.
+    if (instr.bfe.brev) {
+        const auto swap = [&](u32 s, u32 mask) {
+            Node v1 =
+                SignedOperation(OperationCode::ILogicalShiftRight, is_signed, op_a, Immediate(s));
+            if (mask != 0) {
+                v1 = SignedOperation(OperationCode::IBitwiseAnd, is_signed, std::move(v1),
+                                     Immediate(mask));
+            }
+            Node v2 = op_a;
+            if (mask != 0) {
+                v2 = SignedOperation(OperationCode::IBitwiseAnd, is_signed, std::move(v2),
+                                     Immediate(mask));
+            }
+            v2 = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed, std::move(v2),
+                                 Immediate(s));
+            return SignedOperation(OperationCode::IBitwiseOr, is_signed, std::move(v1),
+                                   std::move(v2));
+        };
+        op_a = swap(1, 0x55555555U);
+        op_a = swap(2, 0x33333333U);
+        op_a = swap(4, 0x0F0F0F0FU);
+        op_a = swap(8, 0x00FF00FFU);
+        op_a = swap(16, 0);
     }
 
+    const auto offset = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_b,
+                                        Immediate(0), Immediate(8));
+    const auto bits = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_b,
+                                      Immediate(8), Immediate(8));
+    auto result = SignedOperation(OperationCode::IBitfieldExtract, is_signed, op_a, offset, bits);
+    SetRegister(bb, instr.gpr0, std::move(result));
+
     return pc;
 }
 
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index bee7d8cad..48350e042 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -12,6 +12,7 @@
 #include "common/logging/log.h"
 #include "video_core/engines/shader_bytecode.h"
 #include "video_core/shader/node_helper.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace VideoCommon::Shader {
@@ -359,8 +360,8 @@ ShaderIR::SamplerInfo ShaderIR::GetSamplerInfo(std::optional<SamplerInfo> sample
     if (sampler_info) {
         return *sampler_info;
     }
-    const auto sampler =
-        buffer ? locker.ObtainBindlessSampler(*buffer, offset) : locker.ObtainBoundSampler(offset);
+    const auto sampler = buffer ? registry.ObtainBindlessSampler(*buffer, offset)
+                                : registry.ObtainBoundSampler(offset);
     if (!sampler) {
         LOG_WARNING(HW_GPU, "Unknown sampler info");
         return SamplerInfo{TextureType::Texture2D, false, false, false};
diff --git a/src/video_core/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp
index 206961909..6191ffba1 100644
--- a/src/video_core/shader/decode/xmad.cpp
+++ b/src/video_core/shader/decode/xmad.cpp
@@ -12,6 +12,7 @@ namespace VideoCommon::Shader {
 
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
+using Tegra::Shader::PredCondition;
 
 u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
@@ -30,7 +31,7 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     const bool is_signed_b = instr.xmad.sign_b == 1;
     const bool is_signed_c = is_signed_a;
 
-    auto [is_merge, is_psl, is_high_b, mode, op_b,
+    auto [is_merge, is_psl, is_high_b, mode, op_b_binding,
           op_c] = [&]() -> std::tuple<bool, bool, bool, Tegra::Shader::XmadMode, Node, Node> {
         switch (opcode->get().GetId()) {
         case OpCode::Id::XMAD_CR:
@@ -63,15 +64,19 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
         }
     }();
 
-    op_a = BitfieldExtract(op_a, instr.xmad.high_a ? 16 : 0, 16);
+    op_a = SignedOperation(OperationCode::IBitfieldExtract, is_signed_a, std::move(op_a),
+                           instr.xmad.high_a ? Immediate(16) : Immediate(0), Immediate(16));
 
-    const Node original_b = op_b;
-    op_b = BitfieldExtract(op_b, is_high_b ? 16 : 0, 16);
+    const Node original_b = op_b_binding;
+    const Node op_b =
+        SignedOperation(OperationCode::IBitfieldExtract, is_signed_b, std::move(op_b_binding),
+                        is_high_b ? Immediate(16) : Immediate(0), Immediate(16));
 
-    // TODO(Rodrigo): Use an appropiate sign for this operation
-    Node product = Operation(OperationCode::IMul, NO_PRECISE, op_a, op_b);
+    // we already check sign_a and sign_b is difference or not before so just use one in here.
+    Node product = SignedOperation(OperationCode::IMul, is_signed_a, op_a, op_b);
     if (is_psl) {
-        product = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, product, Immediate(16));
+        product =
+            SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_a, product, Immediate(16));
     }
     SetTemporary(bb, 0, product);
     product = GetTemporary(0);
@@ -88,12 +93,40 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
             return BitfieldExtract(original_c, 16, 16);
         case Tegra::Shader::XmadMode::CBcc: {
             const Node shifted_b = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_b,
-                                                   NO_PRECISE, original_b, Immediate(16));
-            return SignedOperation(OperationCode::IAdd, is_signed_c, NO_PRECISE, original_c,
-                                   shifted_b);
+                                                   original_b, Immediate(16));
+            return SignedOperation(OperationCode::IAdd, is_signed_c, original_c, shifted_b);
+        }
+        case Tegra::Shader::XmadMode::CSfu: {
+            const Node comp_a = GetPredicateComparisonInteger(PredCondition::Equal, is_signed_a,
+                                                              op_a, Immediate(0));
+            const Node comp_b = GetPredicateComparisonInteger(PredCondition::Equal, is_signed_b,
+                                                              op_b, Immediate(0));
+            const Node comp = Operation(OperationCode::LogicalOr, comp_a, comp_b);
+
+            const Node comp_minus_a = GetPredicateComparisonInteger(
+                PredCondition::NotEqual, is_signed_a,
+                SignedOperation(OperationCode::IBitwiseAnd, is_signed_a, op_a,
+                                Immediate(0x80000000)),
+                Immediate(0));
+            const Node comp_minus_b = GetPredicateComparisonInteger(
+                PredCondition::NotEqual, is_signed_b,
+                SignedOperation(OperationCode::IBitwiseAnd, is_signed_b, op_b,
+                                Immediate(0x80000000)),
+                Immediate(0));
+
+            Node new_c = Operation(
+                OperationCode::Select, comp_minus_a,
+                SignedOperation(OperationCode::IAdd, is_signed_c, original_c, Immediate(-65536)),
+                original_c);
+            new_c = Operation(
+                OperationCode::Select, comp_minus_b,
+                SignedOperation(OperationCode::IAdd, is_signed_c, new_c, Immediate(-65536)),
+                std::move(new_c));
+
+            return Operation(OperationCode::Select, comp, original_c, std::move(new_c));
         }
         default:
-            UNIMPLEMENTED_MSG("Unhandled XMAD mode: {}", static_cast<u32>(instr.xmad.mode.Value()));
+            UNREACHABLE();
             return Immediate(0);
         }
     }();
@@ -102,18 +135,19 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
     op_c = GetTemporary(1);
 
     // TODO(Rodrigo): Use an appropiate sign for this operation
-    Node sum = Operation(OperationCode::IAdd, product, op_c);
+    Node sum = SignedOperation(OperationCode::IAdd, is_signed_a, product, std::move(op_c));
     SetTemporary(bb, 2, sum);
     sum = GetTemporary(2);
     if (is_merge) {
-        const Node a = BitfieldExtract(sum, 0, 16);
-        const Node b =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, original_b, Immediate(16));
-        sum = Operation(OperationCode::IBitwiseOr, NO_PRECISE, a, b);
+        const Node a = SignedOperation(OperationCode::IBitfieldExtract, is_signed_a, std::move(sum),
+                                       Immediate(0), Immediate(16));
+        const Node b = SignedOperation(OperationCode::ILogicalShiftLeft, is_signed_b, original_b,
+                                       Immediate(16));
+        sum = SignedOperation(OperationCode::IBitwiseOr, is_signed_a, a, b);
     }
 
     SetInternalFlagsFromInteger(bb, sum, instr.generates_cc);
-    SetRegister(bb, instr.gpr0, sum);
+    SetRegister(bb, instr.gpr0, std::move(sum));
 
     return pc;
 }
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index a0a7b9111..a1828546e 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -299,7 +299,7 @@ private:
     u32 index{};  ///< Emulated index given for the this sampler.
     u32 offset{}; ///< Offset in the const buffer from where the sampler is being read.
     u32 buffer{}; ///< Buffer where the bindless sampler is being read (unused on bound samplers).
-    u32 size{};   ///< Size of the sampler if indexed.
+    u32 size{1};  ///< Size of the sampler.
 
     Tegra::Shader::TextureType type{}; ///< The type used to sample this texture (Texture2D, etc)
     bool is_array{};    ///< Whether the texture is being sampled as an array texture or not.
diff --git a/src/video_core/shader/node_helper.cpp b/src/video_core/shader/node_helper.cpp
index b3dcd291c..76c56abb5 100644
--- a/src/video_core/shader/node_helper.cpp
+++ b/src/video_core/shader/node_helper.cpp
@@ -68,6 +68,8 @@ OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed)
         return OperationCode::UBitwiseXor;
     case OperationCode::IBitwiseNot:
         return OperationCode::UBitwiseNot;
+    case OperationCode::IBitfieldExtract:
+        return OperationCode::UBitfieldExtract;
     case OperationCode::IBitfieldInsert:
         return OperationCode::UBitfieldInsert;
     case OperationCode::IBitCount:
diff --git a/src/video_core/shader/registry.cpp b/src/video_core/shader/registry.cpp
new file mode 100644
index 000000000..af70b3f35
--- /dev/null
+++ b/src/video_core/shader/registry.cpp
@@ -0,0 +1,161 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <tuple>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/kepler_compute.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/engines/shader_type.h"
+#include "video_core/shader/registry.h"
+
+namespace VideoCommon::Shader {
+
+using Tegra::Engines::ConstBufferEngineInterface;
+using Tegra::Engines::SamplerDescriptor;
+using Tegra::Engines::ShaderType;
+
+namespace {
+
+GraphicsInfo MakeGraphicsInfo(ShaderType shader_stage, ConstBufferEngineInterface& engine) {
+    if (shader_stage == ShaderType::Compute) {
+        return {};
+    }
+    auto& graphics = static_cast<Tegra::Engines::Maxwell3D&>(engine);
+
+    GraphicsInfo info;
+    info.tfb_layouts = graphics.regs.tfb_layouts;
+    info.tfb_varying_locs = graphics.regs.tfb_varying_locs;
+    info.primitive_topology = graphics.regs.draw.topology;
+    info.tessellation_primitive = graphics.regs.tess_mode.prim;
+    info.tessellation_spacing = graphics.regs.tess_mode.spacing;
+    info.tfb_enabled = graphics.regs.tfb_enabled;
+    info.tessellation_clockwise = graphics.regs.tess_mode.cw;
+    return info;
+}
+
+ComputeInfo MakeComputeInfo(ShaderType shader_stage, ConstBufferEngineInterface& engine) {
+    if (shader_stage != ShaderType::Compute) {
+        return {};
+    }
+    auto& compute = static_cast<Tegra::Engines::KeplerCompute&>(engine);
+    const auto& launch = compute.launch_description;
+
+    ComputeInfo info;
+    info.workgroup_size = {launch.block_dim_x, launch.block_dim_y, launch.block_dim_z};
+    info.local_memory_size_in_words = launch.local_pos_alloc;
+    info.shared_memory_size_in_words = launch.shared_alloc;
+    return info;
+}
+
+} // Anonymous namespace
+
+Registry::Registry(Tegra::Engines::ShaderType shader_stage, const SerializedRegistryInfo& info)
+    : stage{shader_stage}, stored_guest_driver_profile{info.guest_driver_profile},
+      bound_buffer{info.bound_buffer}, graphics_info{info.graphics}, compute_info{info.compute} {}
+
+Registry::Registry(Tegra::Engines::ShaderType shader_stage,
+                   Tegra::Engines::ConstBufferEngineInterface& engine)
+    : stage{shader_stage}, engine{&engine}, bound_buffer{engine.GetBoundBuffer()},
+      graphics_info{MakeGraphicsInfo(shader_stage, engine)}, compute_info{MakeComputeInfo(
+                                                                 shader_stage, engine)} {}
+
+Registry::~Registry() = default;
+
+std::optional<u32> Registry::ObtainKey(u32 buffer, u32 offset) {
+    const std::pair<u32, u32> key = {buffer, offset};
+    const auto iter = keys.find(key);
+    if (iter != keys.end()) {
+        return iter->second;
+    }
+    if (!engine) {
+        return std::nullopt;
+    }
+    const u32 value = engine->AccessConstBuffer32(stage, buffer, offset);
+    keys.emplace(key, value);
+    return value;
+}
+
+std::optional<SamplerDescriptor> Registry::ObtainBoundSampler(u32 offset) {
+    const u32 key = offset;
+    const auto iter = bound_samplers.find(key);
+    if (iter != bound_samplers.end()) {
+        return iter->second;
+    }
+    if (!engine) {
+        return std::nullopt;
+    }
+    const SamplerDescriptor value = engine->AccessBoundSampler(stage, offset);
+    bound_samplers.emplace(key, value);
+    return value;
+}
+
+std::optional<Tegra::Engines::SamplerDescriptor> Registry::ObtainBindlessSampler(u32 buffer,
+                                                                                 u32 offset) {
+    const std::pair key = {buffer, offset};
+    const auto iter = bindless_samplers.find(key);
+    if (iter != bindless_samplers.end()) {
+        return iter->second;
+    }
+    if (!engine) {
+        return std::nullopt;
+    }
+    const SamplerDescriptor value = engine->AccessBindlessSampler(stage, buffer, offset);
+    bindless_samplers.emplace(key, value);
+    return value;
+}
+
+void Registry::InsertKey(u32 buffer, u32 offset, u32 value) {
+    keys.insert_or_assign({buffer, offset}, value);
+}
+
+void Registry::InsertBoundSampler(u32 offset, SamplerDescriptor sampler) {
+    bound_samplers.insert_or_assign(offset, sampler);
+}
+
+void Registry::InsertBindlessSampler(u32 buffer, u32 offset, SamplerDescriptor sampler) {
+    bindless_samplers.insert_or_assign({buffer, offset}, sampler);
+}
+
+bool Registry::IsConsistent() const {
+    if (!engine) {
+        return true;
+    }
+    return std::all_of(keys.begin(), keys.end(),
+                       [this](const auto& pair) {
+                           const auto [cbuf, offset] = pair.first;
+                           const auto value = pair.second;
+                           return value == engine->AccessConstBuffer32(stage, cbuf, offset);
+                       }) &&
+           std::all_of(bound_samplers.begin(), bound_samplers.end(),
+                       [this](const auto& sampler) {
+                           const auto [key, value] = sampler;
+                           return value == engine->AccessBoundSampler(stage, key);
+                       }) &&
+           std::all_of(bindless_samplers.begin(), bindless_samplers.end(),
+                       [this](const auto& sampler) {
+                           const auto [cbuf, offset] = sampler.first;
+                           const auto value = sampler.second;
+                           return value == engine->AccessBindlessSampler(stage, cbuf, offset);
+                       });
+}
+
+bool Registry::HasEqualKeys(const Registry& rhs) const {
+    return std::tie(keys, bound_samplers, bindless_samplers) ==
+           std::tie(rhs.keys, rhs.bound_samplers, rhs.bindless_samplers);
+}
+
+const GraphicsInfo& Registry::GetGraphicsInfo() const {
+    ASSERT(stage != Tegra::Engines::ShaderType::Compute);
+    return graphics_info;
+}
+
+const ComputeInfo& Registry::GetComputeInfo() const {
+    ASSERT(stage == Tegra::Engines::ShaderType::Compute);
+    return compute_info;
+}
+
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/registry.h b/src/video_core/shader/registry.h
new file mode 100644
index 000000000..0c80d35fd
--- /dev/null
+++ b/src/video_core/shader/registry.h
@@ -0,0 +1,137 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <optional>
+#include <type_traits>
+#include <unordered_map>
+#include <utility>
+
+#include "common/common_types.h"
+#include "common/hash.h"
+#include "video_core/engines/const_buffer_engine_interface.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/engines/shader_type.h"
+#include "video_core/guest_driver.h"
+
+namespace VideoCommon::Shader {
+
+using KeyMap = std::unordered_map<std::pair<u32, u32>, u32, Common::PairHash>;
+using BoundSamplerMap = std::unordered_map<u32, Tegra::Engines::SamplerDescriptor>;
+using BindlessSamplerMap =
+    std::unordered_map<std::pair<u32, u32>, Tegra::Engines::SamplerDescriptor, Common::PairHash>;
+
+struct GraphicsInfo {
+    using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+
+    std::array<Maxwell::TransformFeedbackLayout, Maxwell::NumTransformFeedbackBuffers>
+        tfb_layouts{};
+    std::array<std::array<u8, 128>, Maxwell::NumTransformFeedbackBuffers> tfb_varying_locs{};
+    Maxwell::PrimitiveTopology primitive_topology{};
+    Maxwell::TessellationPrimitive tessellation_primitive{};
+    Maxwell::TessellationSpacing tessellation_spacing{};
+    bool tfb_enabled = false;
+    bool tessellation_clockwise = false;
+};
+static_assert(std::is_trivially_copyable_v<GraphicsInfo> &&
+              std::is_standard_layout_v<GraphicsInfo>);
+
+struct ComputeInfo {
+    std::array<u32, 3> workgroup_size{};
+    u32 shared_memory_size_in_words = 0;
+    u32 local_memory_size_in_words = 0;
+};
+static_assert(std::is_trivially_copyable_v<ComputeInfo> && std::is_standard_layout_v<ComputeInfo>);
+
+struct SerializedRegistryInfo {
+    VideoCore::GuestDriverProfile guest_driver_profile;
+    u32 bound_buffer = 0;
+    GraphicsInfo graphics;
+    ComputeInfo compute;
+};
+
+/**
+ * The Registry is a class use to interface the 3D and compute engines with the shader compiler.
+ * With it, the shader can obtain required data from GPU state and store it for disk shader
+ * compilation.
+ */
+class Registry {
+public:
+    explicit Registry(Tegra::Engines::ShaderType shader_stage, const SerializedRegistryInfo& info);
+
+    explicit Registry(Tegra::Engines::ShaderType shader_stage,
+                      Tegra::Engines::ConstBufferEngineInterface& engine);
+
+    ~Registry();
+
+    /// Retrieves a key from the registry, if it's registered, it will give the registered value, if
+    /// not it will obtain it from maxwell3d and register it.
+    std::optional<u32> ObtainKey(u32 buffer, u32 offset);
+
+    std::optional<Tegra::Engines::SamplerDescriptor> ObtainBoundSampler(u32 offset);
+
+    std::optional<Tegra::Engines::SamplerDescriptor> ObtainBindlessSampler(u32 buffer, u32 offset);
+
+    /// Inserts a key.
+    void InsertKey(u32 buffer, u32 offset, u32 value);
+
+    /// Inserts a bound sampler key.
+    void InsertBoundSampler(u32 offset, Tegra::Engines::SamplerDescriptor sampler);
+
+    /// Inserts a bindless sampler key.
+    void InsertBindlessSampler(u32 buffer, u32 offset, Tegra::Engines::SamplerDescriptor sampler);
+
+    /// Checks keys and samplers against engine's current const buffers.
+    /// Returns true if they are the same value, false otherwise.
+    bool IsConsistent() const;
+
+    /// Returns true if the keys are equal to the other ones in the registry.
+    bool HasEqualKeys(const Registry& rhs) const;
+
+    /// Returns graphics information from this shader
+    const GraphicsInfo& GetGraphicsInfo() const;
+
+    /// Returns compute information from this shader
+    const ComputeInfo& GetComputeInfo() const;
+
+    /// Gives an getter to the const buffer keys in the database.
+    const KeyMap& GetKeys() const {
+        return keys;
+    }
+
+    /// Gets samplers database.
+    const BoundSamplerMap& GetBoundSamplers() const {
+        return bound_samplers;
+    }
+
+    /// Gets bindless samplers database.
+    const BindlessSamplerMap& GetBindlessSamplers() const {
+        return bindless_samplers;
+    }
+
+    /// Gets bound buffer used on this shader
+    u32 GetBoundBuffer() const {
+        return bound_buffer;
+    }
+
+    /// Obtains access to the guest driver's profile.
+    VideoCore::GuestDriverProfile& AccessGuestDriverProfile() {
+        return engine ? engine->AccessGuestDriverProfile() : stored_guest_driver_profile;
+    }
+
+private:
+    const Tegra::Engines::ShaderType stage;
+    VideoCore::GuestDriverProfile stored_guest_driver_profile;
+    Tegra::Engines::ConstBufferEngineInterface* engine = nullptr;
+    KeyMap keys;
+    BoundSamplerMap bound_samplers;
+    BindlessSamplerMap bindless_samplers;
+    u32 bound_buffer;
+    GraphicsInfo graphics_info;
+    ComputeInfo compute_info;
+};
+
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index 3a5d280a9..baf7188d2 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -11,6 +11,7 @@
 #include "common/logging/log.h"
 #include "video_core/engines/shader_bytecode.h"
 #include "video_core/shader/node_helper.h"
+#include "video_core/shader/registry.h"
 #include "video_core/shader/shader_ir.h"
 
 namespace VideoCommon::Shader {
@@ -24,8 +25,8 @@ using Tegra::Shader::PredOperation;
 using Tegra::Shader::Register;
 
 ShaderIR::ShaderIR(const ProgramCode& program_code, u32 main_offset, CompilerSettings settings,
-                   ConstBufferLocker& locker)
-    : program_code{program_code}, main_offset{main_offset}, settings{settings}, locker{locker} {
+                   Registry& registry)
+    : program_code{program_code}, main_offset{main_offset}, settings{settings}, registry{registry} {
     Decode();
     PostDecode();
 }
@@ -95,6 +96,7 @@ Node ShaderIR::GetPredicate(bool immediate) {
 }
 
 Node ShaderIR::GetInputAttribute(Attribute::Index index, u64 element, Node buffer) {
+    MarkAttributeUsage(index, element);
     used_input_attributes.emplace(index);
     return MakeNode<AbufNode>(index, static_cast<u32>(element), std::move(buffer));
 }
@@ -105,42 +107,8 @@ Node ShaderIR::GetPhysicalInputAttribute(Tegra::Shader::Register physical_addres
 }
 
 Node ShaderIR::GetOutputAttribute(Attribute::Index index, u64 element, Node buffer) {
-    if (index == Attribute::Index::LayerViewportPointSize) {
-        switch (element) {
-        case 0:
-            UNIMPLEMENTED();
-            break;
-        case 1:
-            uses_layer = true;
-            break;
-        case 2:
-            uses_viewport_index = true;
-            break;
-        case 3:
-            uses_point_size = true;
-            break;
-        }
-    }
-    if (index == Attribute::Index::TessCoordInstanceIDVertexID) {
-        switch (element) {
-        case 2:
-            uses_instance_id = true;
-            break;
-        case 3:
-            uses_vertex_id = true;
-            break;
-        default:
-            break;
-        }
-    }
-    if (index == Attribute::Index::ClipDistances0123 ||
-        index == Attribute::Index::ClipDistances4567) {
-        const auto clip_index =
-            static_cast<u32>((index == Attribute::Index::ClipDistances4567 ? 1 : 0) + element);
-        used_clip_distances.at(clip_index) = true;
-    }
+    MarkAttributeUsage(index, element);
     used_output_attributes.insert(index);
-
     return MakeNode<AbufNode>(index, static_cast<u32>(element), std::move(buffer));
 }
 
@@ -451,6 +419,54 @@ Node ShaderIR::BitfieldInsert(Node base, Node insert, u32 offset, u32 bits) {
                      Immediate(bits));
 }
 
+void ShaderIR::MarkAttributeUsage(Attribute::Index index, u64 element) {
+    switch (index) {
+    case Attribute::Index::LayerViewportPointSize:
+        switch (element) {
+        case 0:
+            UNIMPLEMENTED();
+            break;
+        case 1:
+            uses_layer = true;
+            break;
+        case 2:
+            uses_viewport_index = true;
+            break;
+        case 3:
+            uses_point_size = true;
+            break;
+        }
+        break;
+    case Attribute::Index::TessCoordInstanceIDVertexID:
+        switch (element) {
+        case 2:
+            uses_instance_id = true;
+            break;
+        case 3:
+            uses_vertex_id = true;
+            break;
+        }
+        break;
+    case Attribute::Index::ClipDistances0123:
+    case Attribute::Index::ClipDistances4567: {
+        const u64 clip_index = (index == Attribute::Index::ClipDistances4567 ? 4 : 0) + element;
+        used_clip_distances.at(clip_index) = true;
+        break;
+    }
+    case Attribute::Index::FrontColor:
+    case Attribute::Index::FrontSecondaryColor:
+    case Attribute::Index::BackColor:
+    case Attribute::Index::BackSecondaryColor:
+        uses_legacy_varyings = true;
+        break;
+    default:
+        if (index >= Attribute::Index::TexCoord_0 && index <= Attribute::Index::TexCoord_7) {
+            uses_legacy_varyings = true;
+        }
+        break;
+    }
+}
+
 std::size_t ShaderIR::DeclareAmend(Node new_amend) {
     const std::size_t id = amend_code.size();
     amend_code.push_back(new_amend);
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index b0851c3be..80fc9b82c 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -18,8 +18,8 @@
 #include "video_core/engines/shader_header.h"
 #include "video_core/shader/ast.h"
 #include "video_core/shader/compiler_settings.h"
-#include "video_core/shader/const_buffer_locker.h"
 #include "video_core/shader/node.h"
+#include "video_core/shader/registry.h"
 
 namespace VideoCommon::Shader {
 
@@ -69,7 +69,7 @@ struct GlobalMemoryUsage {
 class ShaderIR final {
 public:
     explicit ShaderIR(const ProgramCode& program_code, u32 main_offset, CompilerSettings settings,
-                      ConstBufferLocker& locker);
+                      Registry& registry);
     ~ShaderIR();
 
     const std::map<u32, NodeBlock>& GetBasicBlocks() const {
@@ -137,6 +137,10 @@ public:
         return uses_vertex_id;
     }
 
+    bool UsesLegacyVaryings() const {
+        return uses_legacy_varyings;
+    }
+
     bool UsesWarps() const {
         return uses_warps;
     }
@@ -343,6 +347,9 @@ private:
     /// Inserts a sequence of bits from a node
     Node BitfieldInsert(Node base, Node insert, u32 offset, u32 bits);
 
+    /// Marks the usage of a input or output attribute.
+    void MarkAttributeUsage(Tegra::Shader::Attribute::Index index, u64 element);
+
     void WriteTexInstructionFloat(NodeBlock& bb, Tegra::Shader::Instruction instr,
                                   const Node4& components);
 
@@ -414,7 +421,7 @@ private:
     const ProgramCode& program_code;
     const u32 main_offset;
     const CompilerSettings settings;
-    ConstBufferLocker& locker;
+    Registry& registry;
 
     bool decompiled{};
     bool disable_flow_stack{};
@@ -443,6 +450,7 @@ private:
     bool uses_physical_attributes{}; // Shader uses AL2P or physical attribute read/writes
     bool uses_instance_id{};
     bool uses_vertex_id{};
+    bool uses_legacy_varyings{};
     bool uses_warps{};
     bool uses_indexed_samplers{};
 
diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp
index face8c943..10739b37d 100644
--- a/src/video_core/shader/track.cpp
+++ b/src/video_core/shader/track.cpp
@@ -81,26 +81,20 @@ std::tuple<Node, TrackSampler> ShaderIR::TrackBindlessSampler(Node tracked, cons
                 MakeTrackSampler<BindlessSamplerNode>(cbuf->GetIndex(), immediate->GetValue());
             return {tracked, track};
         } else if (const auto operation = std::get_if<OperationNode>(&*offset)) {
-            auto bound_buffer = locker.ObtainBoundBuffer();
-            if (!bound_buffer) {
+            const u32 bound_buffer = registry.GetBoundBuffer();
+            if (bound_buffer != cbuf->GetIndex()) {
                 return {};
             }
-            if (*bound_buffer != cbuf->GetIndex()) {
-                return {};
-            }
-            auto pair = DecoupleIndirectRead(*operation);
+            const auto pair = DecoupleIndirectRead(*operation);
             if (!pair) {
                 return {};
             }
             auto [gpr, base_offset] = *pair;
             const auto offset_inm = std::get_if<ImmediateNode>(&*base_offset);
-            auto gpu_driver = locker.AccessGuestDriverProfile();
-            if (gpu_driver == nullptr) {
-                return {};
-            }
+            const auto& gpu_driver = registry.AccessGuestDriverProfile();
             const u32 bindless_cv = NewCustomVariable();
-            const Node op = Operation(OperationCode::UDiv, NO_PRECISE, gpr,
-                                      Immediate(gpu_driver->GetTextureHandlerSize()));
+            const Node op =
+                Operation(OperationCode::UDiv, gpr, Immediate(gpu_driver.GetTextureHandlerSize()));
 
             const Node cv_node = GetCustomVariable(bindless_cv);
             Node amend_op = Operation(OperationCode::Assign, cv_node, std::move(op));
@@ -157,13 +151,21 @@ std::tuple<Node, u32, u32> ShaderIR::TrackCbuf(Node tracked, const NodeBlock& co
         if (gpr->GetIndex() == Tegra::Shader::Register::ZeroIndex) {
             return {};
         }
-        // Reduce the cursor in one to avoid infinite loops when the instruction sets the same
-        // register that it uses as operand
-        const auto [source, new_cursor] = TrackRegister(gpr, code, cursor - 1);
-        if (!source) {
-            return {};
+        s64 current_cursor = cursor;
+        while (current_cursor > 0) {
+            // Reduce the cursor in one to avoid infinite loops when the instruction sets the same
+            // register that it uses as operand
+            const auto [source, new_cursor] = TrackRegister(gpr, code, current_cursor - 1);
+            current_cursor = new_cursor;
+            if (!source) {
+                continue;
+            }
+            const auto [base_address, index, offset] = TrackCbuf(source, code, current_cursor);
+            if (base_address != nullptr) {
+                return {base_address, index, offset};
+            }
         }
-        return TrackCbuf(source, code, new_cursor);
+        return {};
     }
     if (const auto operation = std::get_if<OperationNode>(&*tracked)) {
         for (std::size_t i = operation->GetOperandsCount(); i > 0; --i) {
diff --git a/src/video_core/shader/transform_feedback.cpp b/src/video_core/shader/transform_feedback.cpp
new file mode 100644
index 000000000..22a933761
--- /dev/null
+++ b/src/video_core/shader/transform_feedback.cpp
@@ -0,0 +1,115 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <unordered_map>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/shader/registry.h"
+#include "video_core/shader/transform_feedback.h"
+
+namespace VideoCommon::Shader {
+
+namespace {
+
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+
+// TODO(Rodrigo): Change this to constexpr std::unordered_set in C++20
+
+/// Attribute offsets that describe a vector
+constexpr std::array VECTORS = {
+    28,  // gl_Position
+    32,  // Generic 0
+    36,  // Generic 1
+    40,  // Generic 2
+    44,  // Generic 3
+    48,  // Generic 4
+    52,  // Generic 5
+    56,  // Generic 6
+    60,  // Generic 7
+    64,  // Generic 8
+    68,  // Generic 9
+    72,  // Generic 10
+    76,  // Generic 11
+    80,  // Generic 12
+    84,  // Generic 13
+    88,  // Generic 14
+    92,  // Generic 15
+    96,  // Generic 16
+    100, // Generic 17
+    104, // Generic 18
+    108, // Generic 19
+    112, // Generic 20
+    116, // Generic 21
+    120, // Generic 22
+    124, // Generic 23
+    128, // Generic 24
+    132, // Generic 25
+    136, // Generic 26
+    140, // Generic 27
+    144, // Generic 28
+    148, // Generic 29
+    152, // Generic 30
+    156, // Generic 31
+    160, // gl_FrontColor
+    164, // gl_FrontSecondaryColor
+    160, // gl_BackColor
+    164, // gl_BackSecondaryColor
+    192, // gl_TexCoord[0]
+    196, // gl_TexCoord[1]
+    200, // gl_TexCoord[2]
+    204, // gl_TexCoord[3]
+    208, // gl_TexCoord[4]
+    212, // gl_TexCoord[5]
+    216, // gl_TexCoord[6]
+    220, // gl_TexCoord[7]
+};
+} // namespace
+
+std::unordered_map<u8, VaryingTFB> BuildTransformFeedback(const GraphicsInfo& info) {
+
+    std::unordered_map<u8, VaryingTFB> tfb;
+
+    for (std::size_t buffer = 0; buffer < Maxwell::NumTransformFeedbackBuffers; ++buffer) {
+        const auto& locations = info.tfb_varying_locs[buffer];
+        const auto& layout = info.tfb_layouts[buffer];
+        const std::size_t varying_count = layout.varying_count;
+
+        std::size_t highest = 0;
+
+        for (std::size_t offset = 0; offset < varying_count; ++offset) {
+            const std::size_t base_offset = offset;
+            const u8 location = locations[offset];
+
+            VaryingTFB varying;
+            varying.buffer = layout.stream;
+            varying.stride = layout.stride;
+            varying.offset = offset * sizeof(u32);
+            varying.components = 1;
+
+            if (std::find(VECTORS.begin(), VECTORS.end(), location / 4 * 4) != VECTORS.end()) {
+                UNIMPLEMENTED_IF_MSG(location % 4 != 0, "Unaligned TFB");
+
+                const u8 base_index = location / 4;
+                while (offset + 1 < varying_count && base_index == locations[offset + 1] / 4) {
+                    ++offset;
+                    ++varying.components;
+                }
+            }
+
+            [[maybe_unused]] const bool inserted = tfb.emplace(location, varying).second;
+            UNIMPLEMENTED_IF_MSG(!inserted, "Varying already stored");
+
+            highest = std::max(highest, (base_offset + varying.components) * sizeof(u32));
+        }
+
+        UNIMPLEMENTED_IF(highest != layout.stride);
+    }
+    return tfb;
+}
+
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/transform_feedback.h b/src/video_core/shader/transform_feedback.h
new file mode 100644
index 000000000..77d05f64c
--- /dev/null
+++ b/src/video_core/shader/transform_feedback.h
@@ -0,0 +1,23 @@
+// Copyright 2020 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/shader/registry.h"
+
+namespace VideoCommon::Shader {
+
+struct VaryingTFB {
+    std::size_t buffer;
+    std::size_t stride;
+    std::size_t offset;
+    std::size_t components;
+};
+
+std::unordered_map<u8, VaryingTFB> BuildTransformFeedback(const GraphicsInfo& info);
+
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp
index 1655ccf16..cc7181229 100644
--- a/src/video_core/surface.cpp
+++ b/src/video_core/surface.cpp
@@ -111,6 +111,8 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format)
         return PixelFormat::RGBA16F;
     case Tegra::RenderTargetFormat::RGBA16_UNORM:
         return PixelFormat::RGBA16U;
+    case Tegra::RenderTargetFormat::RGBA16_SNORM:
+        return PixelFormat::RGBA16S;
     case Tegra::RenderTargetFormat::RGBA16_UINT:
         return PixelFormat::RGBA16UI;
     case Tegra::RenderTargetFormat::RGBA32_FLOAT:
@@ -155,6 +157,8 @@ PixelFormat PixelFormatFromRenderTargetFormat(Tegra::RenderTargetFormat format)
         return PixelFormat::R16I;
     case Tegra::RenderTargetFormat::R32_FLOAT:
         return PixelFormat::R32F;
+    case Tegra::RenderTargetFormat::R32_SINT:
+        return PixelFormat::R32I;
     case Tegra::RenderTargetFormat::R32_UINT:
         return PixelFormat::R32UI;
     case Tegra::RenderTargetFormat::RG32_UINT:
diff --git a/src/video_core/surface.h b/src/video_core/surface.h
index 0d17a93ed..ae8817465 100644
--- a/src/video_core/surface.h
+++ b/src/video_core/surface.h
@@ -25,81 +25,83 @@ enum class PixelFormat {
     R8UI = 7,
     RGBA16F = 8,
     RGBA16U = 9,
-    RGBA16UI = 10,
-    R11FG11FB10F = 11,
-    RGBA32UI = 12,
-    DXT1 = 13,
-    DXT23 = 14,
-    DXT45 = 15,
-    DXN1 = 16, // This is also known as BC4
-    DXN2UNORM = 17,
-    DXN2SNORM = 18,
-    BC7U = 19,
-    BC6H_UF16 = 20,
-    BC6H_SF16 = 21,
-    ASTC_2D_4X4 = 22,
-    BGRA8 = 23,
-    RGBA32F = 24,
-    RG32F = 25,
-    R32F = 26,
-    R16F = 27,
-    R16U = 28,
-    R16S = 29,
-    R16UI = 30,
-    R16I = 31,
-    RG16 = 32,
-    RG16F = 33,
-    RG16UI = 34,
-    RG16I = 35,
-    RG16S = 36,
-    RGB32F = 37,
-    RGBA8_SRGB = 38,
-    RG8U = 39,
-    RG8S = 40,
-    RG32UI = 41,
-    RGBX16F = 42,
-    R32UI = 43,
-    ASTC_2D_8X8 = 44,
-    ASTC_2D_8X5 = 45,
-    ASTC_2D_5X4 = 46,
-    BGRA8_SRGB = 47,
-    DXT1_SRGB = 48,
-    DXT23_SRGB = 49,
-    DXT45_SRGB = 50,
-    BC7U_SRGB = 51,
-    R4G4B4A4U = 52,
-    ASTC_2D_4X4_SRGB = 53,
-    ASTC_2D_8X8_SRGB = 54,
-    ASTC_2D_8X5_SRGB = 55,
-    ASTC_2D_5X4_SRGB = 56,
-    ASTC_2D_5X5 = 57,
-    ASTC_2D_5X5_SRGB = 58,
-    ASTC_2D_10X8 = 59,
-    ASTC_2D_10X8_SRGB = 60,
-    ASTC_2D_6X6 = 61,
-    ASTC_2D_6X6_SRGB = 62,
-    ASTC_2D_10X10 = 63,
-    ASTC_2D_10X10_SRGB = 64,
-    ASTC_2D_12X12 = 65,
-    ASTC_2D_12X12_SRGB = 66,
-    ASTC_2D_8X6 = 67,
-    ASTC_2D_8X6_SRGB = 68,
-    ASTC_2D_6X5 = 69,
-    ASTC_2D_6X5_SRGB = 70,
-    E5B9G9R9F = 71,
+    RGBA16S = 10,
+    RGBA16UI = 11,
+    R11FG11FB10F = 12,
+    RGBA32UI = 13,
+    DXT1 = 14,
+    DXT23 = 15,
+    DXT45 = 16,
+    DXN1 = 17, // This is also known as BC4
+    DXN2UNORM = 18,
+    DXN2SNORM = 19,
+    BC7U = 20,
+    BC6H_UF16 = 21,
+    BC6H_SF16 = 22,
+    ASTC_2D_4X4 = 23,
+    BGRA8 = 24,
+    RGBA32F = 25,
+    RG32F = 26,
+    R32F = 27,
+    R16F = 28,
+    R16U = 29,
+    R16S = 30,
+    R16UI = 31,
+    R16I = 32,
+    RG16 = 33,
+    RG16F = 34,
+    RG16UI = 35,
+    RG16I = 36,
+    RG16S = 37,
+    RGB32F = 38,
+    RGBA8_SRGB = 39,
+    RG8U = 40,
+    RG8S = 41,
+    RG32UI = 42,
+    RGBX16F = 43,
+    R32UI = 44,
+    R32I = 45,
+    ASTC_2D_8X8 = 46,
+    ASTC_2D_8X5 = 47,
+    ASTC_2D_5X4 = 48,
+    BGRA8_SRGB = 49,
+    DXT1_SRGB = 50,
+    DXT23_SRGB = 51,
+    DXT45_SRGB = 52,
+    BC7U_SRGB = 53,
+    R4G4B4A4U = 54,
+    ASTC_2D_4X4_SRGB = 55,
+    ASTC_2D_8X8_SRGB = 56,
+    ASTC_2D_8X5_SRGB = 57,
+    ASTC_2D_5X4_SRGB = 58,
+    ASTC_2D_5X5 = 59,
+    ASTC_2D_5X5_SRGB = 60,
+    ASTC_2D_10X8 = 61,
+    ASTC_2D_10X8_SRGB = 62,
+    ASTC_2D_6X6 = 63,
+    ASTC_2D_6X6_SRGB = 64,
+    ASTC_2D_10X10 = 65,
+    ASTC_2D_10X10_SRGB = 66,
+    ASTC_2D_12X12 = 67,
+    ASTC_2D_12X12_SRGB = 68,
+    ASTC_2D_8X6 = 69,
+    ASTC_2D_8X6_SRGB = 70,
+    ASTC_2D_6X5 = 71,
+    ASTC_2D_6X5_SRGB = 72,
+    E5B9G9R9F = 73,
 
     MaxColorFormat,
 
     // Depth formats
-    Z32F = 72,
-    Z16 = 73,
+    Z32F = 74,
+    Z16 = 75,
 
     MaxDepthFormat,
 
     // DepthStencil formats
-    Z24S8 = 74,
-    S8Z24 = 75,
-    Z32FS8 = 76,
+    Z24S8 = 76,
+    S8Z24 = 77,
+    Z32FS8 = 78,
 
     MaxDepthStencilFormat,
 
@@ -137,6 +139,7 @@ constexpr std::array<u32, MaxPixelFormat> compression_factor_shift_table = {{
     0, // R8UI
     0, // RGBA16F
     0, // RGBA16U
+    0, // RGBA16S
     0, // RGBA16UI
     0, // R11FG11FB10F
     0, // RGBA32UI
@@ -171,6 +174,7 @@ constexpr std::array<u32, MaxPixelFormat> compression_factor_shift_table = {{
     0, // RG32UI
     0, // RGBX16F
     0, // R32UI
+    0, // R32I
     2, // ASTC_2D_8X8
     2, // ASTC_2D_8X5
     2, // ASTC_2D_5X4
@@ -233,6 +237,7 @@ constexpr std::array<u32, MaxPixelFormat> block_width_table = {{
     1,  // R8UI
     1,  // RGBA16F
     1,  // RGBA16U
+    1,  // RGBA16S
     1,  // RGBA16UI
     1,  // R11FG11FB10F
     1,  // RGBA32UI
@@ -267,6 +272,7 @@ constexpr std::array<u32, MaxPixelFormat> block_width_table = {{
     1,  // RG32UI
     1,  // RGBX16F
     1,  // R32UI
+    1,  // R32I
     8,  // ASTC_2D_8X8
     8,  // ASTC_2D_8X5
     5,  // ASTC_2D_5X4
@@ -321,6 +327,7 @@ constexpr std::array<u32, MaxPixelFormat> block_height_table = {{
     1,  // R8UI
     1,  // RGBA16F
     1,  // RGBA16U
+    1,  // RGBA16S
     1,  // RGBA16UI
     1,  // R11FG11FB10F
     1,  // RGBA32UI
@@ -355,6 +362,7 @@ constexpr std::array<u32, MaxPixelFormat> block_height_table = {{
     1,  // RG32UI
     1,  // RGBX16F
     1,  // R32UI
+    1,  // R32I
     8,  // ASTC_2D_8X8
     5,  // ASTC_2D_8X5
     4,  // ASTC_2D_5X4
@@ -409,6 +417,7 @@ constexpr std::array<u32, MaxPixelFormat> bpp_table = {{
     8,   // R8UI
     64,  // RGBA16F
     64,  // RGBA16U
+    64,  // RGBA16S
     64,  // RGBA16UI
     32,  // R11FG11FB10F
     128, // RGBA32UI
@@ -443,6 +452,7 @@ constexpr std::array<u32, MaxPixelFormat> bpp_table = {{
     64,  // RG32UI
     64,  // RGBX16F
     32,  // R32UI
+    32,  // R32I
     128, // ASTC_2D_8X8
     128, // ASTC_2D_8X5
     128, // ASTC_2D_5X4
@@ -512,6 +522,7 @@ constexpr std::array<SurfaceCompression, MaxPixelFormat> compression_type_table
     SurfaceCompression::None,       // R8UI
     SurfaceCompression::None,       // RGBA16F
     SurfaceCompression::None,       // RGBA16U
+    SurfaceCompression::None,       // RGBA16S
     SurfaceCompression::None,       // RGBA16UI
     SurfaceCompression::None,       // R11FG11FB10F
     SurfaceCompression::None,       // RGBA32UI
@@ -546,6 +557,7 @@ constexpr std::array<SurfaceCompression, MaxPixelFormat> compression_type_table
     SurfaceCompression::None,       // RG32UI
     SurfaceCompression::None,       // RGBX16F
     SurfaceCompression::None,       // R32UI
+    SurfaceCompression::None,       // R32I
     SurfaceCompression::Converted,  // ASTC_2D_8X8
     SurfaceCompression::Converted,  // ASTC_2D_8X5
     SurfaceCompression::Converted,  // ASTC_2D_5X4
diff --git a/src/video_core/texture_cache/format_lookup_table.cpp b/src/video_core/texture_cache/format_lookup_table.cpp
index 81fb9f633..e151c26c4 100644
--- a/src/video_core/texture_cache/format_lookup_table.cpp
+++ b/src/video_core/texture_cache/format_lookup_table.cpp
@@ -41,7 +41,7 @@ struct Table {
     ComponentType alpha_component;
     bool is_srgb;
 };
-constexpr std::array<Table, 74> DefinitionTable = {{
+constexpr std::array<Table, 76> DefinitionTable = {{
     {TextureFormat::A8R8G8B8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::ABGR8U},
     {TextureFormat::A8R8G8B8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::ABGR8S},
     {TextureFormat::A8R8G8B8, C, UINT, UINT, UINT, UINT, PixelFormat::ABGR8UI},
@@ -61,6 +61,7 @@ constexpr std::array<Table, 74> DefinitionTable = {{
     {TextureFormat::G8R8, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RG8U},
     {TextureFormat::G8R8, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RG8S},
 
+    {TextureFormat::R16_G16_B16_A16, C, SNORM, SNORM, SNORM, SNORM, PixelFormat::RGBA16S},
     {TextureFormat::R16_G16_B16_A16, C, UNORM, UNORM, UNORM, UNORM, PixelFormat::RGBA16U},
     {TextureFormat::R16_G16_B16_A16, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::RGBA16F},
     {TextureFormat::R16_G16_B16_A16, C, UINT, UINT, UINT, UINT, PixelFormat::RGBA16UI},
@@ -89,6 +90,7 @@ constexpr std::array<Table, 74> DefinitionTable = {{
 
     {TextureFormat::R32, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::R32F},
     {TextureFormat::R32, C, UINT, UINT, UINT, UINT, PixelFormat::R32UI},
+    {TextureFormat::R32, C, SINT, SINT, SINT, SINT, PixelFormat::R32I},
 
     {TextureFormat::E5B9G9R9_SHAREDEXP, C, FLOAT, FLOAT, FLOAT, FLOAT, PixelFormat::E5B9G9R9F},
 
diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp
index 38b3a4ba8..9931c5ef7 100644
--- a/src/video_core/texture_cache/surface_params.cpp
+++ b/src/video_core/texture_cache/surface_params.cpp
@@ -84,19 +84,16 @@ SurfaceParams SurfaceParams::CreateForTexture(const FormatLookupTable& lookup_ta
     if (entry.IsShadow() && params.type == SurfaceType::ColorTexture) {
         switch (params.pixel_format) {
         case PixelFormat::R16U:
-        case PixelFormat::R16F: {
+        case PixelFormat::R16F:
             params.pixel_format = PixelFormat::Z16;
             break;
-        }
-        case PixelFormat::R32F: {
+        case PixelFormat::R32F:
             params.pixel_format = PixelFormat::Z32F;
             break;
-        }
-        default: {
+        default:
             UNIMPLEMENTED_MSG("Unimplemented shadow convert format: {}",
                               static_cast<u32>(params.pixel_format));
         }
-        }
         params.type = GetFormatType(params.pixel_format);
     }
     params.type = GetFormatType(params.pixel_format);
@@ -116,8 +113,10 @@ SurfaceParams SurfaceParams::CreateForTexture(const FormatLookupTable& lookup_ta
         params.height = tic.Height();
         params.depth = tic.Depth();
         params.pitch = params.is_tiled ? 0 : tic.Pitch();
-        if (params.target == SurfaceTarget::TextureCubemap ||
-            params.target == SurfaceTarget::TextureCubeArray) {
+        if (params.target == SurfaceTarget::Texture2D && params.depth > 1) {
+            params.depth = 1;
+        } else if (params.target == SurfaceTarget::TextureCubemap ||
+                   params.target == SurfaceTarget::TextureCubeArray) {
             params.depth *= 6;
         }
         params.num_levels = tic.max_mip_level + 1;
@@ -168,27 +167,29 @@ SurfaceParams SurfaceParams::CreateForImage(const FormatLookupTable& lookup_tabl
     return params;
 }
 
-SurfaceParams SurfaceParams::CreateForDepthBuffer(
-    Core::System& system, u32 zeta_width, u32 zeta_height, Tegra::DepthFormat format,
-    u32 block_width, u32 block_height, u32 block_depth,
-    Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) {
+SurfaceParams SurfaceParams::CreateForDepthBuffer(Core::System& system) {
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    regs.zeta_width, regs.zeta_height, regs.zeta.format, regs.zeta.memory_layout.type;
     SurfaceParams params;
-    params.is_tiled = type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
+    params.is_tiled = regs.zeta.memory_layout.type ==
+                      Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
     params.srgb_conversion = false;
-    params.block_width = std::min(block_width, 5U);
-    params.block_height = std::min(block_height, 5U);
-    params.block_depth = std::min(block_depth, 5U);
+    params.block_width = std::min(regs.zeta.memory_layout.block_width.Value(), 5U);
+    params.block_height = std::min(regs.zeta.memory_layout.block_height.Value(), 5U);
+    params.block_depth = std::min(regs.zeta.memory_layout.block_depth.Value(), 5U);
     params.tile_width_spacing = 1;
-    params.pixel_format = PixelFormatFromDepthFormat(format);
+    params.pixel_format = PixelFormatFromDepthFormat(regs.zeta.format);
     params.type = GetFormatType(params.pixel_format);
-    params.width = zeta_width;
-    params.height = zeta_height;
-    params.target = SurfaceTarget::Texture2D;
-    params.depth = 1;
+    params.width = regs.zeta_width;
+    params.height = regs.zeta_height;
     params.pitch = 0;
     params.num_levels = 1;
     params.emulated_levels = 1;
-    params.is_layered = false;
+
+    const bool is_layered = regs.zeta_layers > 1 && params.block_depth == 0;
+    params.is_layered = is_layered;
+    params.target = is_layered ? SurfaceTarget::Texture2DArray : SurfaceTarget::Texture2D;
+    params.depth = is_layered ? regs.zeta_layers.Value() : 1U;
     return params;
 }
 
@@ -214,11 +215,13 @@ SurfaceParams SurfaceParams::CreateForFramebuffer(Core::System& system, std::siz
         params.width = params.pitch / bpp;
     }
     params.height = config.height;
-    params.depth = 1;
-    params.target = SurfaceTarget::Texture2D;
     params.num_levels = 1;
     params.emulated_levels = 1;
-    params.is_layered = false;
+
+    const bool is_layered = config.layers > 1 && params.block_depth == 0;
+    params.is_layered = is_layered;
+    params.depth = is_layered ? config.layers.Value() : 1;
+    params.target = is_layered ? SurfaceTarget::Texture2DArray : SurfaceTarget::Texture2D;
     return params;
 }
 
diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h
index 9256fd6d9..995cc3818 100644
--- a/src/video_core/texture_cache/surface_params.h
+++ b/src/video_core/texture_cache/surface_params.h
@@ -35,10 +35,7 @@ public:
                                         const VideoCommon::Shader::Image& entry);
 
     /// Creates SurfaceCachedParams for a depth buffer configuration.
-    static SurfaceParams CreateForDepthBuffer(
-        Core::System& system, u32 zeta_width, u32 zeta_height, Tegra::DepthFormat format,
-        u32 block_width, u32 block_height, u32 block_depth,
-        Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type);
+    static SurfaceParams CreateForDepthBuffer(Core::System& system);
 
     /// Creates SurfaceCachedParams from a framebuffer configuration.
     static SurfaceParams CreateForFramebuffer(Core::System& system, std::size_t index);
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 0d105d386..6cdbe63d0 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -22,6 +22,7 @@
 #include "core/core.h"
 #include "core/memory.h"
 #include "core/settings.h"
+#include "video_core/dirty_flags.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
@@ -103,6 +104,11 @@ public:
         if (!cache_addr) {
             return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
         }
+
+        if (!IsTypeCompatible(tic.texture_type, entry)) {
+            return GetNullSurface(SurfaceParams::ExpectedTarget(entry));
+        }
+
         const auto params{SurfaceParams::CreateForTexture(format_lookup_table, tic, entry)};
         const auto [surface, view] = GetSurface(gpu_addr, cache_addr, params, true, false);
         if (guard_samplers) {
@@ -142,11 +148,10 @@ public:
     TView GetDepthBufferSurface(bool preserve_contents) {
         std::lock_guard lock{mutex};
         auto& maxwell3d = system.GPU().Maxwell3D();
-
-        if (!maxwell3d.dirty.depth_buffer) {
+        if (!maxwell3d.dirty.flags[VideoCommon::Dirty::ZetaBuffer]) {
             return depth_buffer.view;
         }
-        maxwell3d.dirty.depth_buffer = false;
+        maxwell3d.dirty.flags[VideoCommon::Dirty::ZetaBuffer] = false;
 
         const auto& regs{maxwell3d.regs};
         const auto gpu_addr{regs.zeta.Address()};
@@ -160,10 +165,7 @@ public:
             SetEmptyDepthBuffer();
             return {};
         }
-        const auto depth_params{SurfaceParams::CreateForDepthBuffer(
-            system, regs.zeta_width, regs.zeta_height, regs.zeta.format,
-            regs.zeta.memory_layout.block_width, regs.zeta.memory_layout.block_height,
-            regs.zeta.memory_layout.block_depth, regs.zeta.memory_layout.type)};
+        const auto depth_params{SurfaceParams::CreateForDepthBuffer(system)};
         auto surface_view = GetSurface(gpu_addr, cache_addr, depth_params, preserve_contents, true);
         if (depth_buffer.target)
             depth_buffer.target->MarkAsRenderTarget(false, NO_RT);
@@ -178,10 +180,10 @@ public:
         std::lock_guard lock{mutex};
         ASSERT(index < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets);
         auto& maxwell3d = system.GPU().Maxwell3D();
-        if (!maxwell3d.dirty.render_target[index]) {
+        if (!maxwell3d.dirty.flags[VideoCommon::Dirty::ColorBuffer0 + index]) {
             return render_targets[index].view;
         }
-        maxwell3d.dirty.render_target[index] = false;
+        maxwell3d.dirty.flags[VideoCommon::Dirty::ColorBuffer0 + index] = false;
 
         const auto& regs{maxwell3d.regs};
         if (index >= regs.rt_control.count || regs.rt[index].Address() == 0 ||
@@ -323,14 +325,14 @@ protected:
     virtual void BufferCopy(TSurface& src_surface, TSurface& dst_surface) = 0;
 
     void ManageRenderTargetUnregister(TSurface& surface) {
-        auto& maxwell3d = system.GPU().Maxwell3D();
+        auto& dirty = system.GPU().Maxwell3D().dirty;
         const u32 index = surface->GetRenderTarget();
         if (index == DEPTH_RT) {
-            maxwell3d.dirty.depth_buffer = true;
+            dirty.flags[VideoCommon::Dirty::ZetaBuffer] = true;
         } else {
-            maxwell3d.dirty.render_target[index] = true;
+            dirty.flags[VideoCommon::Dirty::ColorBuffer0 + index] = true;
         }
-        maxwell3d.dirty.render_settings = true;
+        dirty.flags[VideoCommon::Dirty::RenderTargets] = true;
     }
 
     void Register(TSurface surface) {
@@ -917,13 +919,15 @@ private:
         params.width = 1;
         params.height = 1;
         params.depth = 1;
+        if (target == SurfaceTarget::TextureCubemap || target == SurfaceTarget::TextureCubeArray) {
+            params.depth = 6;
+        }
         params.pitch = 4;
         params.num_levels = 1;
         params.emulated_levels = 1;
-        params.pixel_format = VideoCore::Surface::PixelFormat::RGBA16F;
+        params.pixel_format = VideoCore::Surface::PixelFormat::R8U;
         params.type = VideoCore::Surface::SurfaceType::ColorTexture;
         auto surface = CreateSurface(0ULL, params);
-        invalid_memory.clear();
         invalid_memory.resize(surface->GetHostSizeInBytes(), 0U);
         surface->UploadTexture(invalid_memory);
         surface->MarkAsModified(false, Tick());
@@ -1085,6 +1089,36 @@ private:
         return siblings_table[static_cast<std::size_t>(format)];
     }
 
+    /// Returns true the shader sampler entry is compatible with the TIC texture type.
+    static bool IsTypeCompatible(Tegra::Texture::TextureType tic_type,
+                                 const VideoCommon::Shader::Sampler& entry) {
+        const auto shader_type = entry.GetType();
+        switch (tic_type) {
+        case Tegra::Texture::TextureType::Texture1D:
+        case Tegra::Texture::TextureType::Texture1DArray:
+            return shader_type == Tegra::Shader::TextureType::Texture1D;
+        case Tegra::Texture::TextureType::Texture1DBuffer:
+            // TODO(Rodrigo): Assume as valid for now
+            return true;
+        case Tegra::Texture::TextureType::Texture2D:
+        case Tegra::Texture::TextureType::Texture2DNoMipmap:
+            return shader_type == Tegra::Shader::TextureType::Texture2D;
+        case Tegra::Texture::TextureType::Texture2DArray:
+            return shader_type == Tegra::Shader::TextureType::Texture2D ||
+                   shader_type == Tegra::Shader::TextureType::TextureCube;
+        case Tegra::Texture::TextureType::Texture3D:
+            return shader_type == Tegra::Shader::TextureType::Texture3D;
+        case Tegra::Texture::TextureType::TextureCubeArray:
+        case Tegra::Texture::TextureType::TextureCubemap:
+            if (shader_type == Tegra::Shader::TextureType::TextureCube) {
+                return true;
+            }
+            return shader_type == Tegra::Shader::TextureType::Texture2D && entry.IsArray();
+        }
+        UNREACHABLE();
+        return true;
+    }
+
     struct FramebufferTargetInfo {
         TSurface target;
         TView view;
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp
index 33bd31865..062b4f252 100644
--- a/src/video_core/textures/astc.cpp
+++ b/src/video_core/textures/astc.cpp
@@ -17,26 +17,37 @@
 
 #include <algorithm>
 #include <cassert>
-#include <cstdint>
 #include <cstring>
 #include <vector>
 
+#include "common/common_types.h"
+
 #include "video_core/textures/astc.h"
 
+namespace {
+
+/// Count the number of bits set in a number.
+constexpr u32 Popcnt(u32 n) {
+    u32 c = 0;
+    for (; n; c++) {
+        n &= n - 1;
+    }
+    return c;
+}
+
+} // Anonymous namespace
+
 class InputBitStream {
 public:
-    explicit InputBitStream(const unsigned char* ptr, int start_offset = 0)
+    explicit InputBitStream(const u8* ptr, std::size_t start_offset = 0)
         : m_CurByte(ptr), m_NextBit(start_offset % 8) {}
 
-    ~InputBitStream() = default;
-
-    int GetBitsRead() const {
+    std::size_t GetBitsRead() const {
         return m_BitsRead;
     }
 
-    int ReadBit() {
-
-        int bit = *m_CurByte >> m_NextBit++;
+    u32 ReadBit() {
+        u32 bit = *m_CurByte >> m_NextBit++;
         while (m_NextBit >= 8) {
             m_NextBit -= 8;
             m_CurByte++;
@@ -46,57 +57,66 @@ public:
         return bit & 1;
     }
 
-    unsigned int ReadBits(unsigned int nBits) {
-        unsigned int ret = 0;
-        for (unsigned int i = 0; i < nBits; i++) {
+    u32 ReadBits(std::size_t nBits) {
+        u32 ret = 0;
+        for (std::size_t i = 0; i < nBits; ++i) {
+            ret |= (ReadBit() & 1) << i;
+        }
+        return ret;
+    }
+
+    template <std::size_t nBits>
+    u32 ReadBits() {
+        u32 ret = 0;
+        for (std::size_t i = 0; i < nBits; ++i) {
             ret |= (ReadBit() & 1) << i;
         }
         return ret;
     }
 
 private:
-    const unsigned char* m_CurByte;
-    int m_NextBit = 0;
-    int m_BitsRead = 0;
+    const u8* m_CurByte;
+    std::size_t m_NextBit = 0;
+    std::size_t m_BitsRead = 0;
 };
 
 class OutputBitStream {
 public:
-    explicit OutputBitStream(unsigned char* ptr, int nBits = 0, int start_offset = 0)
+    explicit OutputBitStream(u8* ptr, s32 nBits = 0, s32 start_offset = 0)
         : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {}
 
     ~OutputBitStream() = default;
 
-    int GetBitsWritten() const {
+    s32 GetBitsWritten() const {
         return m_BitsWritten;
     }
 
-    void WriteBitsR(unsigned int val, unsigned int nBits) {
-        for (unsigned int i = 0; i < nBits; i++) {
+    void WriteBitsR(u32 val, u32 nBits) {
+        for (u32 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++) {
+    void WriteBits(u32 val, u32 nBits) {
+        for (u32 i = 0; i < nBits; i++) {
             WriteBit((val >> i) & 1);
         }
     }
 
 private:
-    void WriteBit(int b) {
+    void WriteBit(s32 b) {
 
         if (done)
             return;
 
-        const unsigned int mask = 1 << m_NextBit++;
+        const u32 mask = 1 << m_NextBit++;
 
         // clear the bit
-        *m_CurByte &= static_cast<unsigned char>(~mask);
+        *m_CurByte &= static_cast<u8>(~mask);
 
         // Write the bit, if necessary
         if (b)
-            *m_CurByte |= static_cast<unsigned char>(mask);
+            *m_CurByte |= static_cast<u8>(mask);
 
         // Next byte?
         if (m_NextBit >= 8) {
@@ -107,10 +127,10 @@ private:
         done = done || ++m_BitsWritten >= m_NumBits;
     }
 
-    int m_BitsWritten = 0;
-    const int m_NumBits;
-    unsigned char* m_CurByte;
-    int m_NextBit = 0;
+    s32 m_BitsWritten = 0;
+    const s32 m_NumBits;
+    u8* m_CurByte;
+    s32 m_NextBit = 0;
 
     bool done = false;
 };
@@ -123,20 +143,20 @@ public:
     Bits(const Bits&) = delete;
     Bits& operator=(const Bits&) = delete;
 
-    uint8_t operator[](uint32_t bitPos) const {
-        return static_cast<uint8_t>((m_Bits >> bitPos) & 1);
+    u8 operator[](u32 bitPos) const {
+        return static_cast<u8>((m_Bits >> bitPos) & 1);
     }
 
-    IntType operator()(uint32_t start, uint32_t end) const {
+    IntType operator()(u32 start, u32 end) const {
         if (start == end) {
             return (*this)[start];
         } else if (start > end) {
-            uint32_t t = start;
+            u32 t = start;
             start = end;
             end = t;
         }
 
-        uint64_t mask = (1 << (end - start + 1)) - 1;
+        u64 mask = (1 << (end - start + 1)) - 1;
         return (m_Bits >> start) & static_cast<IntType>(mask);
     }
 
@@ -144,273 +164,236 @@ private:
     const IntType& m_Bits;
 };
 
-enum EIntegerEncoding { eIntegerEncoding_JustBits, eIntegerEncoding_Quint, eIntegerEncoding_Trit };
-
-class IntegerEncodedValue {
-private:
-    const EIntegerEncoding m_Encoding;
-    const uint32_t m_NumBits;
-    uint32_t m_BitValue;
-    union {
-        uint32_t m_QuintValue;
-        uint32_t m_TritValue;
-    };
+enum class IntegerEncoding { JustBits, Qus32, Trit };
 
-public:
-    // Jank, but we're not doing any heavy lifting in this class, so it's
-    // probably OK. It allows us to use these in std::vectors...
-    IntegerEncodedValue& operator=(const IntegerEncodedValue& other) {
-        new (this) IntegerEncodedValue(other);
-        return *this;
-    }
+struct IntegerEncodedValue {
+    constexpr IntegerEncodedValue() = default;
 
-    IntegerEncodedValue(EIntegerEncoding encoding, uint32_t numBits)
-        : m_Encoding(encoding), m_NumBits(numBits) {}
+    constexpr IntegerEncodedValue(IntegerEncoding encoding_, u32 num_bits_)
+        : encoding{encoding_}, num_bits{num_bits_} {}
 
-    EIntegerEncoding GetEncoding() const {
-        return m_Encoding;
-    }
-    uint32_t BaseBitLength() const {
-        return m_NumBits;
-    }
-
-    uint32_t GetBitValue() const {
-        return m_BitValue;
-    }
-    void SetBitValue(uint32_t val) {
-        m_BitValue = val;
-    }
-
-    uint32_t GetTritValue() const {
-        return m_TritValue;
-    }
-    void SetTritValue(uint32_t val) {
-        m_TritValue = val;
-    }
-
-    uint32_t GetQuintValue() const {
-        return m_QuintValue;
-    }
-    void SetQuintValue(uint32_t val) {
-        m_QuintValue = val;
-    }
-
-    bool MatchesEncoding(const IntegerEncodedValue& other) const {
-        return m_Encoding == other.m_Encoding && m_NumBits == other.m_NumBits;
+    constexpr bool MatchesEncoding(const IntegerEncodedValue& other) const {
+        return encoding == other.encoding && num_bits == other.num_bits;
     }
 
     // Returns the number of bits required to encode nVals values.
-    uint32_t GetBitLength(uint32_t nVals) const {
-        uint32_t totalBits = m_NumBits * nVals;
-        if (m_Encoding == eIntegerEncoding_Trit) {
+    u32 GetBitLength(u32 nVals) const {
+        u32 totalBits = num_bits * nVals;
+        if (encoding == IntegerEncoding::Trit) {
             totalBits += (nVals * 8 + 4) / 5;
-        } else if (m_Encoding == eIntegerEncoding_Quint) {
+        } else if (encoding == IntegerEncoding::Qus32) {
             totalBits += (nVals * 7 + 2) / 3;
         }
         return totalBits;
     }
 
-    // Count the number of bits set in a number.
-    static inline uint32_t Popcnt(uint32_t n) {
-        uint32_t c;
-        for (c = 0; n; c++) {
-            n &= n - 1;
+    IntegerEncoding encoding{};
+    u32 num_bits = 0;
+    u32 bit_value = 0;
+    union {
+        u32 qus32_value = 0;
+        u32 trit_value;
+    };
+};
+
+static void DecodeTritBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
+                            u32 nBitsPerValue) {
+    // Implement the algorithm in section C.2.12
+    u32 m[5];
+    u32 t[5];
+    u32 T;
+
+    // Read the trit encoded block according to
+    // table C.2.14
+    m[0] = bits.ReadBits(nBitsPerValue);
+    T = bits.ReadBits<2>();
+    m[1] = bits.ReadBits(nBitsPerValue);
+    T |= bits.ReadBits<2>() << 2;
+    m[2] = bits.ReadBits(nBitsPerValue);
+    T |= bits.ReadBit() << 4;
+    m[3] = bits.ReadBits(nBitsPerValue);
+    T |= bits.ReadBits<2>() << 5;
+    m[4] = bits.ReadBits(nBitsPerValue);
+    T |= bits.ReadBit() << 7;
+
+    u32 C = 0;
+
+    Bits<u32> Tb(T);
+    if (Tb(2, 4) == 7) {
+        C = (Tb(5, 7) << 2) | Tb(0, 1);
+        t[4] = t[3] = 2;
+    } else {
+        C = Tb(0, 4);
+        if (Tb(5, 6) == 3) {
+            t[4] = 2;
+            t[3] = Tb[7];
+        } else {
+            t[4] = Tb[7];
+            t[3] = Tb(5, 6);
         }
-        return c;
     }
 
-    // Returns a new instance of this struct that corresponds to the
-    // can take no more than maxval values
-    static IntegerEncodedValue CreateEncoding(uint32_t maxVal) {
-        while (maxVal > 0) {
-            uint32_t check = maxVal + 1;
-
-            // Is maxVal a power of two?
-            if (!(check & (check - 1))) {
-                return IntegerEncodedValue(eIntegerEncoding_JustBits, Popcnt(maxVal));
-            }
-
-            // Is maxVal of the type 3*2^n - 1?
-            if ((check % 3 == 0) && !((check / 3) & ((check / 3) - 1))) {
-                return IntegerEncodedValue(eIntegerEncoding_Trit, Popcnt(check / 3 - 1));
-            }
+    Bits<u32> Cb(C);
+    if (Cb(0, 1) == 3) {
+        t[2] = 2;
+        t[1] = Cb[4];
+        t[0] = (Cb[3] << 1) | (Cb[2] & ~Cb[3]);
+    } else if (Cb(2, 3) == 3) {
+        t[2] = 2;
+        t[1] = 2;
+        t[0] = Cb(0, 1);
+    } else {
+        t[2] = Cb[4];
+        t[1] = Cb(2, 3);
+        t[0] = (Cb[1] << 1) | (Cb[0] & ~Cb[1]);
+    }
 
-            // Is maxVal of the type 5*2^n - 1?
-            if ((check % 5 == 0) && !((check / 5) & ((check / 5) - 1))) {
-                return IntegerEncodedValue(eIntegerEncoding_Quint, Popcnt(check / 5 - 1));
-            }
+    for (std::size_t i = 0; i < 5; ++i) {
+        IntegerEncodedValue& val = result.emplace_back(IntegerEncoding::Trit, nBitsPerValue);
+        val.bit_value = m[i];
+        val.trit_value = t[i];
+    }
+}
 
-            // Apparently it can't be represented with a bounded integer sequence...
-            // just iterate.
-            maxVal--;
+static void DecodeQus32Block(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
+                             u32 nBitsPerValue) {
+    // Implement the algorithm in section C.2.12
+    u32 m[3];
+    u32 q[3];
+    u32 Q;
+
+    // Read the trit encoded block according to
+    // table C.2.15
+    m[0] = bits.ReadBits(nBitsPerValue);
+    Q = bits.ReadBits<3>();
+    m[1] = bits.ReadBits(nBitsPerValue);
+    Q |= bits.ReadBits<2>() << 3;
+    m[2] = bits.ReadBits(nBitsPerValue);
+    Q |= bits.ReadBits<2>() << 5;
+
+    Bits<u32> Qb(Q);
+    if (Qb(1, 2) == 3 && Qb(5, 6) == 0) {
+        q[0] = q[1] = 4;
+        q[2] = (Qb[0] << 2) | ((Qb[4] & ~Qb[0]) << 1) | (Qb[3] & ~Qb[0]);
+    } else {
+        u32 C = 0;
+        if (Qb(1, 2) == 3) {
+            q[2] = 4;
+            C = (Qb(3, 4) << 3) | ((~Qb(5, 6) & 3) << 1) | Qb[0];
+        } else {
+            q[2] = Qb(5, 6);
+            C = Qb(0, 4);
         }
-        return IntegerEncodedValue(eIntegerEncoding_JustBits, 0);
-    }
-
-    // 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,
-                                      InputBitStream& bits, uint32_t maxRange, uint32_t nValues) {
-        // Determine encoding parameters
-        IntegerEncodedValue val = IntegerEncodedValue::CreateEncoding(maxRange);
-
-        // Start decoding
-        uint32_t nValsDecoded = 0;
-        while (nValsDecoded < nValues) {
-            switch (val.GetEncoding()) {
-            case eIntegerEncoding_Quint:
-                DecodeQuintBlock(bits, result, val.BaseBitLength());
-                nValsDecoded += 3;
-                break;
 
-            case eIntegerEncoding_Trit:
-                DecodeTritBlock(bits, result, val.BaseBitLength());
-                nValsDecoded += 5;
-                break;
-
-            case eIntegerEncoding_JustBits:
-                val.SetBitValue(bits.ReadBits(val.BaseBitLength()));
-                result.push_back(val);
-                nValsDecoded++;
-                break;
-            }
+        Bits<u32> Cb(C);
+        if (Cb(0, 2) == 5) {
+            q[1] = 4;
+            q[0] = Cb(3, 4);
+        } else {
+            q[1] = Cb(3, 4);
+            q[0] = Cb(0, 2);
         }
     }
 
-private:
-    static void DecodeTritBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
-                                uint32_t nBitsPerValue) {
-        // Implement the algorithm in section C.2.12
-        uint32_t m[5];
-        uint32_t t[5];
-        uint32_t T;
-
-        // Read the trit encoded block according to
-        // table C.2.14
-        m[0] = bits.ReadBits(nBitsPerValue);
-        T = bits.ReadBits(2);
-        m[1] = bits.ReadBits(nBitsPerValue);
-        T |= bits.ReadBits(2) << 2;
-        m[2] = bits.ReadBits(nBitsPerValue);
-        T |= bits.ReadBit() << 4;
-        m[3] = bits.ReadBits(nBitsPerValue);
-        T |= bits.ReadBits(2) << 5;
-        m[4] = bits.ReadBits(nBitsPerValue);
-        T |= bits.ReadBit() << 7;
-
-        uint32_t C = 0;
-
-        Bits<uint32_t> Tb(T);
-        if (Tb(2, 4) == 7) {
-            C = (Tb(5, 7) << 2) | Tb(0, 1);
-            t[4] = t[3] = 2;
-        } else {
-            C = Tb(0, 4);
-            if (Tb(5, 6) == 3) {
-                t[4] = 2;
-                t[3] = Tb[7];
-            } else {
-                t[4] = Tb[7];
-                t[3] = Tb(5, 6);
-            }
+    for (std::size_t i = 0; i < 3; ++i) {
+        IntegerEncodedValue& val = result.emplace_back(IntegerEncoding::Qus32, nBitsPerValue);
+        val.bit_value = m[i];
+        val.qus32_value = q[i];
+    }
+}
+
+// Returns a new instance of this struct that corresponds to the
+// can take no more than maxval values
+static constexpr IntegerEncodedValue CreateEncoding(u32 maxVal) {
+    while (maxVal > 0) {
+        u32 check = maxVal + 1;
+
+        // Is maxVal a power of two?
+        if (!(check & (check - 1))) {
+            return IntegerEncodedValue(IntegerEncoding::JustBits, Popcnt(maxVal));
         }
 
-        Bits<uint32_t> Cb(C);
-        if (Cb(0, 1) == 3) {
-            t[2] = 2;
-            t[1] = Cb[4];
-            t[0] = (Cb[3] << 1) | (Cb[2] & ~Cb[3]);
-        } else if (Cb(2, 3) == 3) {
-            t[2] = 2;
-            t[1] = 2;
-            t[0] = Cb(0, 1);
-        } else {
-            t[2] = Cb[4];
-            t[1] = Cb(2, 3);
-            t[0] = (Cb[1] << 1) | (Cb[0] & ~Cb[1]);
+        // Is maxVal of the type 3*2^n - 1?
+        if ((check % 3 == 0) && !((check / 3) & ((check / 3) - 1))) {
+            return IntegerEncodedValue(IntegerEncoding::Trit, Popcnt(check / 3 - 1));
         }
 
-        for (uint32_t i = 0; i < 5; i++) {
-            IntegerEncodedValue val(eIntegerEncoding_Trit, nBitsPerValue);
-            val.SetBitValue(m[i]);
-            val.SetTritValue(t[i]);
-            result.push_back(val);
+        // Is maxVal of the type 5*2^n - 1?
+        if ((check % 5 == 0) && !((check / 5) & ((check / 5) - 1))) {
+            return IntegerEncodedValue(IntegerEncoding::Qus32, Popcnt(check / 5 - 1));
         }
+
+        // Apparently it can't be represented with a bounded integer sequence...
+        // just iterate.
+        maxVal--;
     }
+    return IntegerEncodedValue(IntegerEncoding::JustBits, 0);
+}
 
-    static void DecodeQuintBlock(InputBitStream& bits, std::vector<IntegerEncodedValue>& result,
-                                 uint32_t nBitsPerValue) {
-        // Implement the algorithm in section C.2.12
-        uint32_t m[3];
-        uint32_t q[3];
-        uint32_t Q;
-
-        // Read the trit encoded block according to
-        // table C.2.15
-        m[0] = bits.ReadBits(nBitsPerValue);
-        Q = bits.ReadBits(3);
-        m[1] = bits.ReadBits(nBitsPerValue);
-        Q |= bits.ReadBits(2) << 3;
-        m[2] = bits.ReadBits(nBitsPerValue);
-        Q |= bits.ReadBits(2) << 5;
-
-        Bits<uint32_t> Qb(Q);
-        if (Qb(1, 2) == 3 && Qb(5, 6) == 0) {
-            q[0] = q[1] = 4;
-            q[2] = (Qb[0] << 2) | ((Qb[4] & ~Qb[0]) << 1) | (Qb[3] & ~Qb[0]);
-        } else {
-            uint32_t C = 0;
-            if (Qb(1, 2) == 3) {
-                q[2] = 4;
-                C = (Qb(3, 4) << 3) | ((~Qb(5, 6) & 3) << 1) | Qb[0];
-            } else {
-                q[2] = Qb(5, 6);
-                C = Qb(0, 4);
-            }
+static constexpr std::array<IntegerEncodedValue, 256> MakeEncodedValues() {
+    std::array<IntegerEncodedValue, 256> encodings{};
+    for (std::size_t i = 0; i < encodings.size(); ++i) {
+        encodings[i] = CreateEncoding(static_cast<u32>(i));
+    }
+    return encodings;
+}
 
-            Bits<uint32_t> Cb(C);
-            if (Cb(0, 2) == 5) {
-                q[1] = 4;
-                q[0] = Cb(3, 4);
-            } else {
-                q[1] = Cb(3, 4);
-                q[0] = Cb(0, 2);
-            }
-        }
+static constexpr std::array EncodingsValues = MakeEncodedValues();
+
+// 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, InputBitStream& bits,
+                                  u32 maxRange, u32 nValues) {
+    // Determine encoding parameters
+    IntegerEncodedValue val = EncodingsValues[maxRange];
+
+    // Start decoding
+    u32 nValsDecoded = 0;
+    while (nValsDecoded < nValues) {
+        switch (val.encoding) {
+        case IntegerEncoding::Qus32:
+            DecodeQus32Block(bits, result, val.num_bits);
+            nValsDecoded += 3;
+            break;
+
+        case IntegerEncoding::Trit:
+            DecodeTritBlock(bits, result, val.num_bits);
+            nValsDecoded += 5;
+            break;
 
-        for (uint32_t i = 0; i < 3; i++) {
-            IntegerEncodedValue val(eIntegerEncoding_Quint, nBitsPerValue);
-            val.m_BitValue = m[i];
-            val.m_QuintValue = q[i];
+        case IntegerEncoding::JustBits:
+            val.bit_value = bits.ReadBits(val.num_bits);
             result.push_back(val);
+            nValsDecoded++;
+            break;
         }
     }
-};
+}
 
 namespace ASTCC {
 
 struct TexelWeightParams {
-    uint32_t m_Width = 0;
-    uint32_t m_Height = 0;
+    u32 m_Width = 0;
+    u32 m_Height = 0;
     bool m_bDualPlane = false;
-    uint32_t m_MaxWeight = 0;
+    u32 m_MaxWeight = 0;
     bool m_bError = false;
     bool m_bVoidExtentLDR = false;
     bool m_bVoidExtentHDR = false;
 
-    uint32_t GetPackedBitSize() const {
+    u32 GetPackedBitSize() const {
         // How many indices do we have?
-        uint32_t nIdxs = m_Height * m_Width;
+        u32 nIdxs = m_Height * m_Width;
         if (m_bDualPlane) {
             nIdxs *= 2;
         }
 
-        return IntegerEncodedValue::CreateEncoding(m_MaxWeight).GetBitLength(nIdxs);
+        return EncodingsValues[m_MaxWeight].GetBitLength(nIdxs);
     }
 
-    uint32_t GetNumWeightValues() const {
-        uint32_t ret = m_Width * m_Height;
+    u32 GetNumWeightValues() const {
+        u32 ret = m_Width * m_Height;
         if (m_bDualPlane) {
             ret *= 2;
         }
@@ -422,7 +405,7 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     TexelWeightParams params;
 
     // Read the entire block mode all at once
-    uint16_t modeBits = static_cast<uint16_t>(strm.ReadBits(11));
+    u16 modeBits = static_cast<u16>(strm.ReadBits<11>());
 
     // Does this match the void extent block mode?
     if ((modeBits & 0x01FF) == 0x1FC) {
@@ -457,7 +440,7 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     // of the block mode. Layout is determined by a number
     // between 0 and 9 corresponding to table C.2.8 of the
     // ASTC spec.
-    uint32_t layout = 0;
+    u32 layout = 0;
 
     if ((modeBits & 0x1) || (modeBits & 0x2)) {
         // layout is in [0-4]
@@ -509,7 +492,7 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     assert(layout < 10);
 
     // Determine R
-    uint32_t R = !!(modeBits & 0x10);
+    u32 R = !!(modeBits & 0x10);
     if (layout < 5) {
         R |= (modeBits & 0x3) << 1;
     } else {
@@ -520,54 +503,54 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     // Determine width & height
     switch (layout) {
     case 0: {
-        uint32_t A = (modeBits >> 5) & 0x3;
-        uint32_t B = (modeBits >> 7) & 0x3;
+        u32 A = (modeBits >> 5) & 0x3;
+        u32 B = (modeBits >> 7) & 0x3;
         params.m_Width = B + 4;
         params.m_Height = A + 2;
         break;
     }
 
     case 1: {
-        uint32_t A = (modeBits >> 5) & 0x3;
-        uint32_t B = (modeBits >> 7) & 0x3;
+        u32 A = (modeBits >> 5) & 0x3;
+        u32 B = (modeBits >> 7) & 0x3;
         params.m_Width = B + 8;
         params.m_Height = A + 2;
         break;
     }
 
     case 2: {
-        uint32_t A = (modeBits >> 5) & 0x3;
-        uint32_t B = (modeBits >> 7) & 0x3;
+        u32 A = (modeBits >> 5) & 0x3;
+        u32 B = (modeBits >> 7) & 0x3;
         params.m_Width = A + 2;
         params.m_Height = B + 8;
         break;
     }
 
     case 3: {
-        uint32_t A = (modeBits >> 5) & 0x3;
-        uint32_t B = (modeBits >> 7) & 0x1;
+        u32 A = (modeBits >> 5) & 0x3;
+        u32 B = (modeBits >> 7) & 0x1;
         params.m_Width = A + 2;
         params.m_Height = B + 6;
         break;
     }
 
     case 4: {
-        uint32_t A = (modeBits >> 5) & 0x3;
-        uint32_t B = (modeBits >> 7) & 0x1;
+        u32 A = (modeBits >> 5) & 0x3;
+        u32 B = (modeBits >> 7) & 0x1;
         params.m_Width = B + 2;
         params.m_Height = A + 2;
         break;
     }
 
     case 5: {
-        uint32_t A = (modeBits >> 5) & 0x3;
+        u32 A = (modeBits >> 5) & 0x3;
         params.m_Width = 12;
         params.m_Height = A + 2;
         break;
     }
 
     case 6: {
-        uint32_t A = (modeBits >> 5) & 0x3;
+        u32 A = (modeBits >> 5) & 0x3;
         params.m_Width = A + 2;
         params.m_Height = 12;
         break;
@@ -586,15 +569,15 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     }
 
     case 9: {
-        uint32_t A = (modeBits >> 5) & 0x3;
-        uint32_t B = (modeBits >> 9) & 0x3;
+        u32 A = (modeBits >> 5) & 0x3;
+        u32 B = (modeBits >> 9) & 0x3;
         params.m_Width = A + 6;
         params.m_Height = B + 6;
         break;
     }
 
     default:
-        assert(!"Don't know this layout...");
+        assert(false && "Don't know this layout...");
         params.m_bError = true;
         break;
     }
@@ -605,10 +588,10 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     bool H = (layout != 9) && (modeBits & 0x200);
 
     if (H) {
-        const uint32_t maxWeights[6] = {9, 11, 15, 19, 23, 31};
+        const u32 maxWeights[6] = {9, 11, 15, 19, 23, 31};
         params.m_MaxWeight = maxWeights[R - 2];
     } else {
-        const uint32_t maxWeights[6] = {1, 2, 3, 4, 5, 7};
+        const u32 maxWeights[6] = {1, 2, 3, 4, 5, 7};
         params.m_MaxWeight = maxWeights[R - 2];
     }
 
@@ -617,32 +600,32 @@ static TexelWeightParams DecodeBlockInfo(InputBitStream& strm) {
     return params;
 }
 
-static void FillVoidExtentLDR(InputBitStream& strm, uint32_t* const outBuf, uint32_t blockWidth,
-                              uint32_t blockHeight) {
+static void FillVoidExtentLDR(InputBitStream& strm, u32* const outBuf, u32 blockWidth,
+                              u32 blockHeight) {
     // Don't actually care about the void extent, just read the bits...
-    for (int i = 0; i < 4; ++i) {
-        strm.ReadBits(13);
+    for (s32 i = 0; i < 4; ++i) {
+        strm.ReadBits<13>();
     }
 
     // Decode the RGBA components and renormalize them to the range [0, 255]
-    uint16_t r = static_cast<uint16_t>(strm.ReadBits(16));
-    uint16_t g = static_cast<uint16_t>(strm.ReadBits(16));
-    uint16_t b = static_cast<uint16_t>(strm.ReadBits(16));
-    uint16_t a = static_cast<uint16_t>(strm.ReadBits(16));
+    u16 r = static_cast<u16>(strm.ReadBits<16>());
+    u16 g = static_cast<u16>(strm.ReadBits<16>());
+    u16 b = static_cast<u16>(strm.ReadBits<16>());
+    u16 a = static_cast<u16>(strm.ReadBits<16>());
 
-    uint32_t rgba = (r >> 8) | (g & 0xFF00) | (static_cast<uint32_t>(b) & 0xFF00) << 8 |
-                    (static_cast<uint32_t>(a) & 0xFF00) << 16;
+    u32 rgba = (r >> 8) | (g & 0xFF00) | (static_cast<u32>(b) & 0xFF00) << 8 |
+               (static_cast<u32>(a) & 0xFF00) << 16;
 
-    for (uint32_t j = 0; j < blockHeight; j++) {
-        for (uint32_t i = 0; i < blockWidth; i++) {
+    for (u32 j = 0; j < blockHeight; j++) {
+        for (u32 i = 0; i < blockWidth; i++) {
             outBuf[j * blockWidth + i] = rgba;
         }
     }
 }
 
-static void FillError(uint32_t* outBuf, uint32_t blockWidth, uint32_t blockHeight) {
-    for (uint32_t j = 0; j < blockHeight; j++) {
-        for (uint32_t i = 0; i < blockWidth; i++) {
+static void FillError(u32* outBuf, u32 blockWidth, u32 blockHeight) {
+    for (u32 j = 0; j < blockHeight; j++) {
+        for (u32 i = 0; i < blockWidth; i++) {
             outBuf[j * blockWidth + i] = 0xFFFF00FF;
         }
     }
@@ -651,18 +634,18 @@ static void FillError(uint32_t* outBuf, uint32_t blockWidth, uint32_t blockHeigh
 // Replicates low numBits such that [(toBit - 1):(toBit - 1 - fromBit)]
 // is the same as [(numBits - 1):0] and repeats all the way down.
 template <typename IntType>
-static IntType Replicate(const IntType& val, uint32_t numBits, uint32_t toBit) {
+static IntType Replicate(IntType val, u32 numBits, u32 toBit) {
     if (numBits == 0)
         return 0;
     if (toBit == 0)
         return 0;
     IntType v = val & static_cast<IntType>((1 << numBits) - 1);
     IntType res = v;
-    uint32_t reslen = numBits;
+    u32 reslen = numBits;
     while (reslen < toBit) {
-        uint32_t comp = 0;
+        u32 comp = 0;
         if (numBits > toBit - reslen) {
-            uint32_t newshift = toBit - reslen;
+            u32 newshift = toBit - reslen;
             comp = numBits - newshift;
             numBits = newshift;
         }
@@ -675,14 +658,14 @@ static IntType Replicate(const IntType& val, uint32_t numBits, uint32_t toBit) {
 
 class Pixel {
 protected:
-    using ChannelType = int16_t;
-    uint8_t m_BitDepth[4] = {8, 8, 8, 8};
-    int16_t color[4] = {};
+    using ChannelType = s16;
+    u8 m_BitDepth[4] = {8, 8, 8, 8};
+    s16 color[4] = {};
 
 public:
     Pixel() = default;
-    Pixel(uint32_t a, uint32_t r, uint32_t g, uint32_t b, unsigned bitDepth = 8)
-        : m_BitDepth{uint8_t(bitDepth), uint8_t(bitDepth), uint8_t(bitDepth), uint8_t(bitDepth)},
+    Pixel(u32 a, u32 r, u32 g, u32 b, u32 bitDepth = 8)
+        : m_BitDepth{u8(bitDepth), u8(bitDepth), u8(bitDepth), u8(bitDepth)},
           color{static_cast<ChannelType>(a), static_cast<ChannelType>(r),
                 static_cast<ChannelType>(g), static_cast<ChannelType>(b)} {}
 
@@ -691,22 +674,22 @@ public:
     // significant bits when going from larger to smaller bit depth
     // or by repeating the most significant bits when going from
     // smaller to larger bit depths.
-    void ChangeBitDepth(const uint8_t (&depth)[4]) {
-        for (uint32_t i = 0; i < 4; i++) {
+    void ChangeBitDepth(const u8 (&depth)[4]) {
+        for (u32 i = 0; i < 4; i++) {
             Component(i) = ChangeBitDepth(Component(i), m_BitDepth[i], depth[i]);
             m_BitDepth[i] = depth[i];
         }
     }
 
     template <typename IntType>
-    static float ConvertChannelToFloat(IntType channel, uint8_t bitDepth) {
+    static float ConvertChannelToFloat(IntType channel, u8 bitDepth) {
         float denominator = static_cast<float>((1 << bitDepth) - 1);
         return static_cast<float>(channel) / denominator;
     }
 
     // Changes the bit depth of a single component. See the comment
     // above for how we do this.
-    static ChannelType ChangeBitDepth(Pixel::ChannelType val, uint8_t oldDepth, uint8_t newDepth) {
+    static ChannelType ChangeBitDepth(Pixel::ChannelType val, u8 oldDepth, u8 newDepth) {
         assert(newDepth <= 8);
         assert(oldDepth <= 8);
 
@@ -722,16 +705,15 @@ public:
             if (newDepth == 0) {
                 return 0xFF;
             } else {
-                uint8_t bitsWasted = static_cast<uint8_t>(oldDepth - newDepth);
-                uint16_t v = static_cast<uint16_t>(val);
-                v = static_cast<uint16_t>((v + (1 << (bitsWasted - 1))) >> bitsWasted);
-                v = ::std::min<uint16_t>(::std::max<uint16_t>(0, v),
-                                         static_cast<uint16_t>((1 << newDepth) - 1));
-                return static_cast<uint8_t>(v);
+                u8 bitsWasted = static_cast<u8>(oldDepth - newDepth);
+                u16 v = static_cast<u16>(val);
+                v = static_cast<u16>((v + (1 << (bitsWasted - 1))) >> bitsWasted);
+                v = ::std::min<u16>(::std::max<u16>(0, v), static_cast<u16>((1 << newDepth) - 1));
+                return static_cast<u8>(v);
             }
         }
 
-        assert(!"We shouldn't get here.");
+        assert(false && "We shouldn't get here.");
         return 0;
     }
 
@@ -759,15 +741,15 @@ public:
     ChannelType& B() {
         return color[3];
     }
-    const ChannelType& Component(uint32_t idx) const {
+    const ChannelType& Component(u32 idx) const {
         return color[idx];
     }
-    ChannelType& Component(uint32_t idx) {
+    ChannelType& Component(u32 idx) {
         return color[idx];
     }
 
-    void GetBitDepth(uint8_t (&outDepth)[4]) const {
-        for (int i = 0; i < 4; i++) {
+    void GetBitDepth(u8 (&outDepth)[4]) const {
+        for (s32 i = 0; i < 4; i++) {
             outDepth[i] = m_BitDepth[i];
         }
     }
@@ -776,12 +758,12 @@ public:
     // and then pack each channel into an R8G8B8A8 32-bit integer. We assume
     // that the architecture is little-endian, so the alpha channel will end
     // up in the most-significant byte.
-    uint32_t Pack() const {
+    u32 Pack() const {
         Pixel eightBit(*this);
-        const uint8_t eightBitDepth[4] = {8, 8, 8, 8};
+        const u8 eightBitDepth[4] = {8, 8, 8, 8};
         eightBit.ChangeBitDepth(eightBitDepth);
 
-        uint32_t r = 0;
+        u32 r = 0;
         r |= eightBit.A();
         r <<= 8;
         r |= eightBit.B();
@@ -794,7 +776,7 @@ public:
 
     // Clamps the pixel to the range [0,255]
     void ClampByte() {
-        for (uint32_t i = 0; i < 4; i++) {
+        for (u32 i = 0; i < 4; i++) {
             color[i] = (color[i] < 0) ? 0 : ((color[i] > 255) ? 255 : color[i]);
         }
     }
@@ -804,24 +786,24 @@ public:
     }
 };
 
-static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* modes,
-                              const uint32_t nPartitions, const uint32_t nBitsForColorData) {
+static void DecodeColorValues(u32* out, u8* data, const u32* modes, const u32 nPartitions,
+                              const u32 nBitsForColorData) {
     // First figure out how many color values we have
-    uint32_t nValues = 0;
-    for (uint32_t i = 0; i < nPartitions; i++) {
+    u32 nValues = 0;
+    for (u32 i = 0; i < nPartitions; i++) {
         nValues += ((modes[i] >> 2) + 1) << 1;
     }
 
     // Then based on the number of values and the remaining number of bits,
     // figure out the max value for each of them...
-    uint32_t range = 256;
+    u32 range = 256;
     while (--range > 0) {
-        IntegerEncodedValue val = IntegerEncodedValue::CreateEncoding(range);
-        uint32_t bitLength = val.GetBitLength(nValues);
+        IntegerEncodedValue val = EncodingsValues[range];
+        u32 bitLength = val.GetBitLength(nValues);
         if (bitLength <= nBitsForColorData) {
             // Find the smallest possible range that matches the given encoding
             while (--range > 0) {
-                IntegerEncodedValue newval = IntegerEncodedValue::CreateEncoding(range);
+                IntegerEncodedValue newval = EncodingsValues[range];
                 if (!newval.MatchesEncoding(val)) {
                     break;
                 }
@@ -835,12 +817,14 @@ 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;
+    decodedColorValues.reserve(32);
+
     InputBitStream colorStream(data);
-    IntegerEncodedValue::DecodeIntegerSequence(decodedColorValues, colorStream, range, nValues);
+    DecodeIntegerSequence(decodedColorValues, colorStream, range, nValues);
 
     // Once we have the decoded values, we need to dequantize them to the 0-255 range
     // This procedure is outlined in ASTC spec C.2.13
-    uint32_t outIdx = 0;
+    u32 outIdx = 0;
     for (auto itr = decodedColorValues.begin(); itr != decodedColorValues.end(); ++itr) {
         // Have we already decoded all that we need?
         if (outIdx >= nValues) {
@@ -848,25 +832,25 @@ static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* mode
         }
 
         const IntegerEncodedValue& val = *itr;
-        uint32_t bitlen = val.BaseBitLength();
-        uint32_t bitval = val.GetBitValue();
+        u32 bitlen = val.num_bits;
+        u32 bitval = val.bit_value;
 
         assert(bitlen >= 1);
 
-        uint32_t A = 0, B = 0, C = 0, D = 0;
+        u32 A = 0, B = 0, C = 0, D = 0;
         // A is just the lsb replicated 9 times.
         A = Replicate(bitval & 1, 1, 9);
 
-        switch (val.GetEncoding()) {
+        switch (val.encoding) {
         // Replicate bits
-        case eIntegerEncoding_JustBits:
+        case IntegerEncoding::JustBits:
             out[outIdx++] = Replicate(bitval, bitlen, 8);
             break;
 
         // Use algorithm in C.2.13
-        case eIntegerEncoding_Trit: {
+        case IntegerEncoding::Trit: {
 
-            D = val.GetTritValue();
+            D = val.trit_value;
 
             switch (bitlen) {
             case 1: {
@@ -876,48 +860,48 @@ static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* mode
             case 2: {
                 C = 93;
                 // B = b000b0bb0
-                uint32_t b = (bitval >> 1) & 1;
+                u32 b = (bitval >> 1) & 1;
                 B = (b << 8) | (b << 4) | (b << 2) | (b << 1);
             } break;
 
             case 3: {
                 C = 44;
                 // B = cb000cbcb
-                uint32_t cb = (bitval >> 1) & 3;
+                u32 cb = (bitval >> 1) & 3;
                 B = (cb << 7) | (cb << 2) | cb;
             } break;
 
             case 4: {
                 C = 22;
                 // B = dcb000dcb
-                uint32_t dcb = (bitval >> 1) & 7;
+                u32 dcb = (bitval >> 1) & 7;
                 B = (dcb << 6) | dcb;
             } break;
 
             case 5: {
                 C = 11;
                 // B = edcb000ed
-                uint32_t edcb = (bitval >> 1) & 0xF;
+                u32 edcb = (bitval >> 1) & 0xF;
                 B = (edcb << 5) | (edcb >> 2);
             } break;
 
             case 6: {
                 C = 5;
                 // B = fedcb000f
-                uint32_t fedcb = (bitval >> 1) & 0x1F;
+                u32 fedcb = (bitval >> 1) & 0x1F;
                 B = (fedcb << 4) | (fedcb >> 4);
             } break;
 
             default:
-                assert(!"Unsupported trit encoding for color values!");
+                assert(false && "Unsupported trit encoding for color values!");
                 break;
             } // switch(bitlen)
-        }     // case eIntegerEncoding_Trit
+        }     // case IntegerEncoding::Trit
         break;
 
-        case eIntegerEncoding_Quint: {
+        case IntegerEncoding::Qus32: {
 
-            D = val.GetQuintValue();
+            D = val.qus32_value;
 
             switch (bitlen) {
             case 1: {
@@ -927,41 +911,41 @@ static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* mode
             case 2: {
                 C = 54;
                 // B = b0000bb00
-                uint32_t b = (bitval >> 1) & 1;
+                u32 b = (bitval >> 1) & 1;
                 B = (b << 8) | (b << 3) | (b << 2);
             } break;
 
             case 3: {
                 C = 26;
                 // B = cb0000cbc
-                uint32_t cb = (bitval >> 1) & 3;
+                u32 cb = (bitval >> 1) & 3;
                 B = (cb << 7) | (cb << 1) | (cb >> 1);
             } break;
 
             case 4: {
                 C = 13;
                 // B = dcb0000dc
-                uint32_t dcb = (bitval >> 1) & 7;
+                u32 dcb = (bitval >> 1) & 7;
                 B = (dcb << 6) | (dcb >> 1);
             } break;
 
             case 5: {
                 C = 6;
                 // B = edcb0000e
-                uint32_t edcb = (bitval >> 1) & 0xF;
+                u32 edcb = (bitval >> 1) & 0xF;
                 B = (edcb << 5) | (edcb >> 3);
             } break;
 
             default:
-                assert(!"Unsupported quint encoding for color values!");
+                assert(false && "Unsupported quint encoding for color values!");
                 break;
             } // switch(bitlen)
-        }     // case eIntegerEncoding_Quint
+        }     // case IntegerEncoding::Qus32
         break;
-        } // switch(val.GetEncoding())
+        } // switch(val.encoding)
 
-        if (val.GetEncoding() != eIntegerEncoding_JustBits) {
-            uint32_t T = D * C + B;
+        if (val.encoding != IntegerEncoding::JustBits) {
+            u32 T = D * C + B;
             T ^= A;
             T = (A & 0x80) | (T >> 2);
             out[outIdx++] = T;
@@ -969,31 +953,31 @@ static void DecodeColorValues(uint32_t* out, uint8_t* data, const uint32_t* mode
     }
 
     // Make sure that each of our values is in the proper range...
-    for (uint32_t i = 0; i < nValues; i++) {
+    for (u32 i = 0; i < nValues; i++) {
         assert(out[i] <= 255);
     }
 }
 
-static uint32_t UnquantizeTexelWeight(const IntegerEncodedValue& val) {
-    uint32_t bitval = val.GetBitValue();
-    uint32_t bitlen = val.BaseBitLength();
+static u32 UnquantizeTexelWeight(const IntegerEncodedValue& val) {
+    u32 bitval = val.bit_value;
+    u32 bitlen = val.num_bits;
 
-    uint32_t A = Replicate(bitval & 1, 1, 7);
-    uint32_t B = 0, C = 0, D = 0;
+    u32 A = Replicate(bitval & 1, 1, 7);
+    u32 B = 0, C = 0, D = 0;
 
-    uint32_t result = 0;
-    switch (val.GetEncoding()) {
-    case eIntegerEncoding_JustBits:
+    u32 result = 0;
+    switch (val.encoding) {
+    case IntegerEncoding::JustBits:
         result = Replicate(bitval, bitlen, 6);
         break;
 
-    case eIntegerEncoding_Trit: {
-        D = val.GetTritValue();
+    case IntegerEncoding::Trit: {
+        D = val.trit_value;
         assert(D < 3);
 
         switch (bitlen) {
         case 0: {
-            uint32_t results[3] = {0, 32, 63};
+            u32 results[3] = {0, 32, 63};
             result = results[D];
         } break;
 
@@ -1003,29 +987,29 @@ static uint32_t UnquantizeTexelWeight(const IntegerEncodedValue& val) {
 
         case 2: {
             C = 23;
-            uint32_t b = (bitval >> 1) & 1;
+            u32 b = (bitval >> 1) & 1;
             B = (b << 6) | (b << 2) | b;
         } break;
 
         case 3: {
             C = 11;
-            uint32_t cb = (bitval >> 1) & 3;
+            u32 cb = (bitval >> 1) & 3;
             B = (cb << 5) | cb;
         } break;
 
         default:
-            assert(!"Invalid trit encoding for texel weight");
+            assert(false && "Invalid trit encoding for texel weight");
             break;
         }
     } break;
 
-    case eIntegerEncoding_Quint: {
-        D = val.GetQuintValue();
+    case IntegerEncoding::Qus32: {
+        D = val.qus32_value;
         assert(D < 5);
 
         switch (bitlen) {
         case 0: {
-            uint32_t results[5] = {0, 16, 32, 47, 63};
+            u32 results[5] = {0, 16, 32, 47, 63};
             result = results[D];
         } break;
 
@@ -1035,18 +1019,18 @@ static uint32_t UnquantizeTexelWeight(const IntegerEncodedValue& val) {
 
         case 2: {
             C = 13;
-            uint32_t b = (bitval >> 1) & 1;
+            u32 b = (bitval >> 1) & 1;
             B = (b << 6) | (b << 1);
         } break;
 
         default:
-            assert(!"Invalid quint encoding for texel weight");
+            assert(false && "Invalid quint encoding for texel weight");
             break;
         }
     } break;
     }
 
-    if (val.GetEncoding() != eIntegerEncoding_JustBits && bitlen > 0) {
+    if (val.encoding != IntegerEncoding::JustBits && bitlen > 0) {
         // Decode the value...
         result = D * C + B;
         result ^= A;
@@ -1063,12 +1047,11 @@ static uint32_t UnquantizeTexelWeight(const IntegerEncodedValue& val) {
     return result;
 }
 
-static void UnquantizeTexelWeights(uint32_t out[2][144],
-                                   const std::vector<IntegerEncodedValue>& weights,
-                                   const TexelWeightParams& params, const uint32_t blockWidth,
-                                   const uint32_t blockHeight) {
-    uint32_t weightIdx = 0;
-    uint32_t unquantized[2][144];
+static void UnquantizeTexelWeights(u32 out[2][144], const std::vector<IntegerEncodedValue>& weights,
+                                   const TexelWeightParams& params, const u32 blockWidth,
+                                   const u32 blockHeight) {
+    u32 weightIdx = 0;
+    u32 unquantized[2][144];
 
     for (auto itr = weights.begin(); itr != weights.end(); ++itr) {
         unquantized[0][weightIdx] = UnquantizeTexelWeight(*itr);
@@ -1086,34 +1069,34 @@ static void UnquantizeTexelWeights(uint32_t out[2][144],
     }
 
     // Do infill if necessary (Section C.2.18) ...
-    uint32_t Ds = (1024 + (blockWidth / 2)) / (blockWidth - 1);
-    uint32_t Dt = (1024 + (blockHeight / 2)) / (blockHeight - 1);
+    u32 Ds = (1024 + (blockWidth / 2)) / (blockWidth - 1);
+    u32 Dt = (1024 + (blockHeight / 2)) / (blockHeight - 1);
 
-    const uint32_t kPlaneScale = params.m_bDualPlane ? 2U : 1U;
-    for (uint32_t plane = 0; plane < kPlaneScale; plane++)
-        for (uint32_t t = 0; t < blockHeight; t++)
-            for (uint32_t s = 0; s < blockWidth; s++) {
-                uint32_t cs = Ds * s;
-                uint32_t ct = Dt * t;
+    const u32 kPlaneScale = params.m_bDualPlane ? 2U : 1U;
+    for (u32 plane = 0; plane < kPlaneScale; plane++)
+        for (u32 t = 0; t < blockHeight; t++)
+            for (u32 s = 0; s < blockWidth; s++) {
+                u32 cs = Ds * s;
+                u32 ct = Dt * t;
 
-                uint32_t gs = (cs * (params.m_Width - 1) + 32) >> 6;
-                uint32_t gt = (ct * (params.m_Height - 1) + 32) >> 6;
+                u32 gs = (cs * (params.m_Width - 1) + 32) >> 6;
+                u32 gt = (ct * (params.m_Height - 1) + 32) >> 6;
 
-                uint32_t js = gs >> 4;
-                uint32_t fs = gs & 0xF;
+                u32 js = gs >> 4;
+                u32 fs = gs & 0xF;
 
-                uint32_t jt = gt >> 4;
-                uint32_t ft = gt & 0x0F;
+                u32 jt = gt >> 4;
+                u32 ft = gt & 0x0F;
 
-                uint32_t w11 = (fs * ft + 8) >> 4;
-                uint32_t w10 = ft - w11;
-                uint32_t w01 = fs - w11;
-                uint32_t w00 = 16 - fs - ft + w11;
+                u32 w11 = (fs * ft + 8) >> 4;
+                u32 w10 = ft - w11;
+                u32 w01 = fs - w11;
+                u32 w00 = 16 - fs - ft + w11;
 
-                uint32_t v0 = js + jt * params.m_Width;
+                u32 v0 = js + jt * params.m_Width;
 
 #define FIND_TEXEL(tidx, bidx)                                                                     \
-    uint32_t p##bidx = 0;                                                                          \
+    u32 p##bidx = 0;                                                                               \
     do {                                                                                           \
         if ((tidx) < (params.m_Width * params.m_Height)) {                                         \
             p##bidx = unquantized[plane][(tidx)];                                                  \
@@ -1133,7 +1116,7 @@ static void UnquantizeTexelWeights(uint32_t out[2][144],
 }
 
 // Transfers a bit as described in C.2.14
-static inline void BitTransferSigned(int32_t& a, int32_t& b) {
+static inline void BitTransferSigned(s32& a, s32& b) {
     b >>= 1;
     b |= a & 0x80;
     a >>= 1;
@@ -1144,14 +1127,14 @@ static inline void BitTransferSigned(int32_t& a, int32_t& b) {
 
 // Adds more precision to the blue channel as described
 // in C.2.14
-static inline Pixel BlueContract(int32_t a, int32_t r, int32_t g, int32_t b) {
-    return Pixel(static_cast<int16_t>(a), static_cast<int16_t>((r + b) >> 1),
-                 static_cast<int16_t>((g + b) >> 1), static_cast<int16_t>(b));
+static inline Pixel BlueContract(s32 a, s32 r, s32 g, s32 b) {
+    return Pixel(static_cast<s16>(a), static_cast<s16>((r + b) >> 1),
+                 static_cast<s16>((g + b) >> 1), static_cast<s16>(b));
 }
 
 // Partition selection functions as specified in
 // C.2.21
-static inline uint32_t hash52(uint32_t p) {
+static inline u32 hash52(u32 p) {
     p ^= p >> 15;
     p -= p << 17;
     p += p << 7;
@@ -1165,8 +1148,7 @@ static inline uint32_t hash52(uint32_t p) {
     return p;
 }
 
-static uint32_t SelectPartition(int32_t seed, int32_t x, int32_t y, int32_t z,
-                                int32_t partitionCount, int32_t smallBlock) {
+static u32 SelectPartition(s32 seed, s32 x, s32 y, s32 z, s32 partitionCount, s32 smallBlock) {
     if (1 == partitionCount)
         return 0;
 
@@ -1178,34 +1160,34 @@ static uint32_t SelectPartition(int32_t seed, int32_t x, int32_t y, int32_t z,
 
     seed += (partitionCount - 1) * 1024;
 
-    uint32_t rnum = hash52(static_cast<uint32_t>(seed));
-    uint8_t seed1 = static_cast<uint8_t>(rnum & 0xF);
-    uint8_t seed2 = static_cast<uint8_t>((rnum >> 4) & 0xF);
-    uint8_t seed3 = static_cast<uint8_t>((rnum >> 8) & 0xF);
-    uint8_t seed4 = static_cast<uint8_t>((rnum >> 12) & 0xF);
-    uint8_t seed5 = static_cast<uint8_t>((rnum >> 16) & 0xF);
-    uint8_t seed6 = static_cast<uint8_t>((rnum >> 20) & 0xF);
-    uint8_t seed7 = static_cast<uint8_t>((rnum >> 24) & 0xF);
-    uint8_t seed8 = static_cast<uint8_t>((rnum >> 28) & 0xF);
-    uint8_t seed9 = static_cast<uint8_t>((rnum >> 18) & 0xF);
-    uint8_t seed10 = static_cast<uint8_t>((rnum >> 22) & 0xF);
-    uint8_t seed11 = static_cast<uint8_t>((rnum >> 26) & 0xF);
-    uint8_t seed12 = static_cast<uint8_t>(((rnum >> 30) | (rnum << 2)) & 0xF);
-
-    seed1 = static_cast<uint8_t>(seed1 * seed1);
-    seed2 = static_cast<uint8_t>(seed2 * seed2);
-    seed3 = static_cast<uint8_t>(seed3 * seed3);
-    seed4 = static_cast<uint8_t>(seed4 * seed4);
-    seed5 = static_cast<uint8_t>(seed5 * seed5);
-    seed6 = static_cast<uint8_t>(seed6 * seed6);
-    seed7 = static_cast<uint8_t>(seed7 * seed7);
-    seed8 = static_cast<uint8_t>(seed8 * seed8);
-    seed9 = static_cast<uint8_t>(seed9 * seed9);
-    seed10 = static_cast<uint8_t>(seed10 * seed10);
-    seed11 = static_cast<uint8_t>(seed11 * seed11);
-    seed12 = static_cast<uint8_t>(seed12 * seed12);
-
-    int32_t sh1, sh2, sh3;
+    u32 rnum = hash52(static_cast<u32>(seed));
+    u8 seed1 = static_cast<u8>(rnum & 0xF);
+    u8 seed2 = static_cast<u8>((rnum >> 4) & 0xF);
+    u8 seed3 = static_cast<u8>((rnum >> 8) & 0xF);
+    u8 seed4 = static_cast<u8>((rnum >> 12) & 0xF);
+    u8 seed5 = static_cast<u8>((rnum >> 16) & 0xF);
+    u8 seed6 = static_cast<u8>((rnum >> 20) & 0xF);
+    u8 seed7 = static_cast<u8>((rnum >> 24) & 0xF);
+    u8 seed8 = static_cast<u8>((rnum >> 28) & 0xF);
+    u8 seed9 = static_cast<u8>((rnum >> 18) & 0xF);
+    u8 seed10 = static_cast<u8>((rnum >> 22) & 0xF);
+    u8 seed11 = static_cast<u8>((rnum >> 26) & 0xF);
+    u8 seed12 = static_cast<u8>(((rnum >> 30) | (rnum << 2)) & 0xF);
+
+    seed1 = static_cast<u8>(seed1 * seed1);
+    seed2 = static_cast<u8>(seed2 * seed2);
+    seed3 = static_cast<u8>(seed3 * seed3);
+    seed4 = static_cast<u8>(seed4 * seed4);
+    seed5 = static_cast<u8>(seed5 * seed5);
+    seed6 = static_cast<u8>(seed6 * seed6);
+    seed7 = static_cast<u8>(seed7 * seed7);
+    seed8 = static_cast<u8>(seed8 * seed8);
+    seed9 = static_cast<u8>(seed9 * seed9);
+    seed10 = static_cast<u8>(seed10 * seed10);
+    seed11 = static_cast<u8>(seed11 * seed11);
+    seed12 = static_cast<u8>(seed12 * seed12);
+
+    s32 sh1, sh2, sh3;
     if (seed & 1) {
         sh1 = (seed & 2) ? 4 : 5;
         sh2 = (partitionCount == 3) ? 6 : 5;
@@ -1215,23 +1197,23 @@ static uint32_t SelectPartition(int32_t seed, int32_t x, int32_t y, int32_t z,
     }
     sh3 = (seed & 0x10) ? sh1 : sh2;
 
-    seed1 = static_cast<uint8_t>(seed1 >> sh1);
-    seed2 = static_cast<uint8_t>(seed2 >> sh2);
-    seed3 = static_cast<uint8_t>(seed3 >> sh1);
-    seed4 = static_cast<uint8_t>(seed4 >> sh2);
-    seed5 = static_cast<uint8_t>(seed5 >> sh1);
-    seed6 = static_cast<uint8_t>(seed6 >> sh2);
-    seed7 = static_cast<uint8_t>(seed7 >> sh1);
-    seed8 = static_cast<uint8_t>(seed8 >> sh2);
-    seed9 = static_cast<uint8_t>(seed9 >> sh3);
-    seed10 = static_cast<uint8_t>(seed10 >> sh3);
-    seed11 = static_cast<uint8_t>(seed11 >> sh3);
-    seed12 = static_cast<uint8_t>(seed12 >> sh3);
-
-    int32_t a = seed1 * x + seed2 * y + seed11 * z + (rnum >> 14);
-    int32_t b = seed3 * x + seed4 * y + seed12 * z + (rnum >> 10);
-    int32_t c = seed5 * x + seed6 * y + seed9 * z + (rnum >> 6);
-    int32_t d = seed7 * x + seed8 * y + seed10 * z + (rnum >> 2);
+    seed1 = static_cast<u8>(seed1 >> sh1);
+    seed2 = static_cast<u8>(seed2 >> sh2);
+    seed3 = static_cast<u8>(seed3 >> sh1);
+    seed4 = static_cast<u8>(seed4 >> sh2);
+    seed5 = static_cast<u8>(seed5 >> sh1);
+    seed6 = static_cast<u8>(seed6 >> sh2);
+    seed7 = static_cast<u8>(seed7 >> sh1);
+    seed8 = static_cast<u8>(seed8 >> sh2);
+    seed9 = static_cast<u8>(seed9 >> sh3);
+    seed10 = static_cast<u8>(seed10 >> sh3);
+    seed11 = static_cast<u8>(seed11 >> sh3);
+    seed12 = static_cast<u8>(seed12 >> sh3);
+
+    s32 a = seed1 * x + seed2 * y + seed11 * z + (rnum >> 14);
+    s32 b = seed3 * x + seed4 * y + seed12 * z + (rnum >> 10);
+    s32 c = seed5 * x + seed6 * y + seed9 * z + (rnum >> 6);
+    s32 d = seed7 * x + seed8 * y + seed10 * z + (rnum >> 2);
 
     a &= 0x3F;
     b &= 0x3F;
@@ -1252,27 +1234,26 @@ static uint32_t SelectPartition(int32_t seed, int32_t x, int32_t y, int32_t z,
     return 3;
 }
 
-static inline uint32_t Select2DPartition(int32_t seed, int32_t x, int32_t y, int32_t partitionCount,
-                                         int32_t smallBlock) {
+static inline u32 Select2DPartition(s32 seed, s32 x, s32 y, s32 partitionCount, s32 smallBlock) {
     return SelectPartition(seed, x, y, 0, partitionCount, smallBlock);
 }
 
 // Section C.2.14
-static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValues,
-                             uint32_t colorEndpointMode) {
+static void ComputeEndpos32s(Pixel& ep1, Pixel& ep2, const u32*& colorValues,
+                             u32 colorEndpos32Mode) {
 #define READ_UINT_VALUES(N)                                                                        \
-    uint32_t v[N];                                                                                 \
-    for (uint32_t i = 0; i < N; i++) {                                                             \
+    u32 v[N];                                                                                      \
+    for (u32 i = 0; i < N; i++) {                                                                  \
         v[i] = *(colorValues++);                                                                   \
     }
 
 #define READ_INT_VALUES(N)                                                                         \
-    int32_t v[N];                                                                                  \
-    for (uint32_t i = 0; i < N; i++) {                                                             \
-        v[i] = static_cast<int32_t>(*(colorValues++));                                             \
+    s32 v[N];                                                                                      \
+    for (u32 i = 0; i < N; i++) {                                                                  \
+        v[i] = static_cast<s32>(*(colorValues++));                                                 \
     }
 
-    switch (colorEndpointMode) {
+    switch (colorEndpos32Mode) {
     case 0: {
         READ_UINT_VALUES(2)
         ep1 = Pixel(0xFF, v[0], v[0], v[0]);
@@ -1281,8 +1262,8 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValue
 
     case 1: {
         READ_UINT_VALUES(2)
-        uint32_t L0 = (v[0] >> 2) | (v[1] & 0xC0);
-        uint32_t L1 = std::max(L0 + (v[1] & 0x3F), 0xFFU);
+        u32 L0 = (v[0] >> 2) | (v[1] & 0xC0);
+        u32 L1 = std::max(L0 + (v[1] & 0x3F), 0xFFU);
         ep1 = Pixel(0xFF, L0, L0, L0);
         ep2 = Pixel(0xFF, L1, L1, L1);
     } break;
@@ -1371,7 +1352,7 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValue
     } break;
 
     default:
-        assert(!"Unsupported color endpoint mode (is it HDR?)");
+        assert(false && "Unsupported color endpoint mode (is it HDR?)");
         break;
     }
 
@@ -1379,14 +1360,14 @@ static void ComputeEndpoints(Pixel& ep1, Pixel& ep2, const uint32_t*& colorValue
 #undef READ_INT_VALUES
 }
 
-static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
-                            const uint32_t blockHeight, uint32_t* outBuf) {
+static void DecompressBlock(const u8 inBuf[16], const u32 blockWidth, const u32 blockHeight,
+                            u32* outBuf) {
     InputBitStream strm(inBuf);
     TexelWeightParams weightParams = DecodeBlockInfo(strm);
 
     // Was there an error?
     if (weightParams.m_bError) {
-        assert(!"Invalid block mode");
+        assert(false && "Invalid block mode");
         FillError(outBuf, blockWidth, blockHeight);
         return;
     }
@@ -1397,63 +1378,63 @@ static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
     }
 
     if (weightParams.m_bVoidExtentHDR) {
-        assert(!"HDR void extent blocks are unsupported!");
+        assert(false && "HDR void extent blocks are unsupported!");
         FillError(outBuf, blockWidth, blockHeight);
         return;
     }
 
     if (weightParams.m_Width > blockWidth) {
-        assert(!"Texel weight grid width should be smaller than block width");
+        assert(false && "Texel weight grid width should be smaller than block width");
         FillError(outBuf, blockWidth, blockHeight);
         return;
     }
 
     if (weightParams.m_Height > blockHeight) {
-        assert(!"Texel weight grid height should be smaller than block height");
+        assert(false && "Texel weight grid height should be smaller than block height");
         FillError(outBuf, blockWidth, blockHeight);
         return;
     }
 
     // Read num partitions
-    uint32_t nPartitions = strm.ReadBits(2) + 1;
+    u32 nPartitions = strm.ReadBits<2>() + 1;
     assert(nPartitions <= 4);
 
     if (nPartitions == 4 && weightParams.m_bDualPlane) {
-        assert(!"Dual plane mode is incompatible with four partition blocks");
+        assert(false && "Dual plane mode is incompatible with four partition blocks");
         FillError(outBuf, blockWidth, blockHeight);
         return;
     }
 
-    // Based on the number of partitions, read the color endpoint mode for
+    // Based on the number of partitions, read the color endpos32 mode for
     // each partition.
 
-    // Determine partitions, partition index, and color endpoint modes
-    int32_t planeIdx = -1;
-    uint32_t partitionIndex;
-    uint32_t colorEndpointMode[4] = {0, 0, 0, 0};
+    // Determine partitions, partition index, and color endpos32 modes
+    s32 planeIdx = -1;
+    u32 partitionIndex;
+    u32 colorEndpos32Mode[4] = {0, 0, 0, 0};
 
     // Define color data.
-    uint8_t colorEndpointData[16];
-    memset(colorEndpointData, 0, sizeof(colorEndpointData));
-    OutputBitStream colorEndpointStream(colorEndpointData, 16 * 8, 0);
+    u8 colorEndpos32Data[16];
+    memset(colorEndpos32Data, 0, sizeof(colorEndpos32Data));
+    OutputBitStream colorEndpos32Stream(colorEndpos32Data, 16 * 8, 0);
 
     // Read extra config data...
-    uint32_t baseCEM = 0;
+    u32 baseCEM = 0;
     if (nPartitions == 1) {
-        colorEndpointMode[0] = strm.ReadBits(4);
+        colorEndpos32Mode[0] = strm.ReadBits<4>();
         partitionIndex = 0;
     } else {
-        partitionIndex = strm.ReadBits(10);
-        baseCEM = strm.ReadBits(6);
+        partitionIndex = strm.ReadBits<10>();
+        baseCEM = strm.ReadBits<6>();
     }
-    uint32_t baseMode = (baseCEM & 3);
+    u32 baseMode = (baseCEM & 3);
 
-    // Remaining bits are color endpoint data...
-    uint32_t nWeightBits = weightParams.GetPackedBitSize();
-    int32_t remainingBits = 128 - nWeightBits - strm.GetBitsRead();
+    // Remaining bits are color endpos32 data...
+    u32 nWeightBits = weightParams.GetPackedBitSize();
+    s32 remainingBits = 128 - nWeightBits - static_cast<s32>(strm.GetBitsRead());
 
     // Consider extra bits prior to texel data...
-    uint32_t extraCEMbits = 0;
+    u32 extraCEMbits = 0;
     if (baseMode) {
         switch (nPartitions) {
         case 2:
@@ -1473,18 +1454,18 @@ static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
     remainingBits -= extraCEMbits;
 
     // Do we have a dual plane situation?
-    uint32_t planeSelectorBits = 0;
+    u32 planeSelectorBits = 0;
     if (weightParams.m_bDualPlane) {
         planeSelectorBits = 2;
     }
     remainingBits -= planeSelectorBits;
 
     // Read color data...
-    uint32_t colorDataBits = remainingBits;
+    u32 colorDataBits = remainingBits;
     while (remainingBits > 0) {
-        uint32_t nb = std::min(remainingBits, 8);
-        uint32_t b = strm.ReadBits(nb);
-        colorEndpointStream.WriteBits(b, nb);
+        u32 nb = std::min(remainingBits, 8);
+        u32 b = strm.ReadBits(nb);
+        colorEndpos32Stream.WriteBits(b, nb);
         remainingBits -= 8;
     }
 
@@ -1493,64 +1474,64 @@ static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
 
     // Read the rest of the CEM
     if (baseMode) {
-        uint32_t extraCEM = strm.ReadBits(extraCEMbits);
-        uint32_t CEM = (extraCEM << 6) | baseCEM;
+        u32 extraCEM = strm.ReadBits(extraCEMbits);
+        u32 CEM = (extraCEM << 6) | baseCEM;
         CEM >>= 2;
 
         bool C[4] = {0};
-        for (uint32_t i = 0; i < nPartitions; i++) {
+        for (u32 i = 0; i < nPartitions; i++) {
             C[i] = CEM & 1;
             CEM >>= 1;
         }
 
-        uint8_t M[4] = {0};
-        for (uint32_t i = 0; i < nPartitions; i++) {
+        u8 M[4] = {0};
+        for (u32 i = 0; i < nPartitions; i++) {
             M[i] = CEM & 3;
             CEM >>= 2;
             assert(M[i] <= 3);
         }
 
-        for (uint32_t i = 0; i < nPartitions; i++) {
-            colorEndpointMode[i] = baseMode;
+        for (u32 i = 0; i < nPartitions; i++) {
+            colorEndpos32Mode[i] = baseMode;
             if (!(C[i]))
-                colorEndpointMode[i] -= 1;
-            colorEndpointMode[i] <<= 2;
-            colorEndpointMode[i] |= M[i];
+                colorEndpos32Mode[i] -= 1;
+            colorEndpos32Mode[i] <<= 2;
+            colorEndpos32Mode[i] |= M[i];
         }
     } else if (nPartitions > 1) {
-        uint32_t CEM = baseCEM >> 2;
-        for (uint32_t i = 0; i < nPartitions; i++) {
-            colorEndpointMode[i] = CEM;
+        u32 CEM = baseCEM >> 2;
+        for (u32 i = 0; i < nPartitions; i++) {
+            colorEndpos32Mode[i] = CEM;
         }
     }
 
     // Make sure everything up till here is sane.
-    for (uint32_t i = 0; i < nPartitions; i++) {
-        assert(colorEndpointMode[i] < 16);
+    for (u32 i = 0; i < nPartitions; i++) {
+        assert(colorEndpos32Mode[i] < 16);
     }
     assert(strm.GetBitsRead() + weightParams.GetPackedBitSize() == 128);
 
     // Decode both color data and texel weight data
-    uint32_t colorValues[32]; // Four values, two endpoints, four maximum paritions
-    DecodeColorValues(colorValues, colorEndpointData, colorEndpointMode, nPartitions,
+    u32 colorValues[32]; // Four values, two endpos32s, four maximum paritions
+    DecodeColorValues(colorValues, colorEndpos32Data, colorEndpos32Mode, nPartitions,
                       colorDataBits);
 
-    Pixel endpoints[4][2];
-    const uint32_t* colorValuesPtr = colorValues;
-    for (uint32_t i = 0; i < nPartitions; i++) {
-        ComputeEndpoints(endpoints[i][0], endpoints[i][1], colorValuesPtr, colorEndpointMode[i]);
+    Pixel endpos32s[4][2];
+    const u32* colorValuesPtr = colorValues;
+    for (u32 i = 0; i < nPartitions; i++) {
+        ComputeEndpos32s(endpos32s[i][0], endpos32s[i][1], colorValuesPtr, colorEndpos32Mode[i]);
     }
 
     // Read the texel weight data..
-    uint8_t texelWeightData[16];
+    u8 texelWeightData[16];
     memcpy(texelWeightData, inBuf, sizeof(texelWeightData));
 
     // Reverse everything
-    for (uint32_t i = 0; i < 8; i++) {
+    for (u32 i = 0; i < 8; i++) {
 // Taken from http://graphics.stanford.edu/~seander/bithacks.html#ReverseByteWith64Bits
 #define REVERSE_BYTE(b) (((b)*0x80200802ULL) & 0x0884422110ULL) * 0x0101010101ULL >> 32
-        unsigned char a = static_cast<unsigned char>(REVERSE_BYTE(texelWeightData[i]));
-        unsigned char b = static_cast<unsigned char>(REVERSE_BYTE(texelWeightData[15 - i]));
+        u8 a = static_cast<u8>(REVERSE_BYTE(texelWeightData[i]));
+        u8 b = static_cast<u8>(REVERSE_BYTE(texelWeightData[15 - i]));
 #undef REVERSE_BYTE
 
         texelWeightData[i] = b;
@@ -1558,50 +1539,51 @@ static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
     }
 
     // Make sure that higher non-texel bits are set to zero
-    const uint32_t clearByteStart = (weightParams.GetPackedBitSize() >> 3) + 1;
+    const u32 clearByteStart = (weightParams.GetPackedBitSize() >> 3) + 1;
     texelWeightData[clearByteStart - 1] =
         texelWeightData[clearByteStart - 1] &
-        static_cast<uint8_t>((1 << (weightParams.GetPackedBitSize() % 8)) - 1);
+        static_cast<u8>((1 << (weightParams.GetPackedBitSize() % 8)) - 1);
     memset(texelWeightData + clearByteStart, 0, 16 - clearByteStart);
 
     std::vector<IntegerEncodedValue> texelWeightValues;
+    texelWeightValues.reserve(64);
+
     InputBitStream weightStream(texelWeightData);
 
-    IntegerEncodedValue::DecodeIntegerSequence(texelWeightValues, weightStream,
-                                               weightParams.m_MaxWeight,
-                                               weightParams.GetNumWeightValues());
+    DecodeIntegerSequence(texelWeightValues, weightStream, weightParams.m_MaxWeight,
+                          weightParams.GetNumWeightValues());
 
     // Blocks can be at most 12x12, so we can have as many as 144 weights
-    uint32_t weights[2][144];
+    u32 weights[2][144];
     UnquantizeTexelWeights(weights, texelWeightValues, weightParams, blockWidth, blockHeight);
 
-    // Now that we have endpoints and weights, we can interpolate and generate
+    // Now that we have endpos32s and weights, we can s32erpolate and generate
     // the proper decoding...
-    for (uint32_t j = 0; j < blockHeight; j++)
-        for (uint32_t i = 0; i < blockWidth; i++) {
-            uint32_t partition = Select2DPartition(partitionIndex, i, j, nPartitions,
-                                                   (blockHeight * blockWidth) < 32);
+    for (u32 j = 0; j < blockHeight; j++)
+        for (u32 i = 0; i < blockWidth; i++) {
+            u32 partition = Select2DPartition(partitionIndex, i, j, nPartitions,
+                                              (blockHeight * blockWidth) < 32);
             assert(partition < nPartitions);
 
             Pixel p;
-            for (uint32_t c = 0; c < 4; c++) {
-                uint32_t C0 = endpoints[partition][0].Component(c);
+            for (u32 c = 0; c < 4; c++) {
+                u32 C0 = endpos32s[partition][0].Component(c);
                 C0 = Replicate(C0, 8, 16);
-                uint32_t C1 = endpoints[partition][1].Component(c);
+                u32 C1 = endpos32s[partition][1].Component(c);
                 C1 = Replicate(C1, 8, 16);
 
-                uint32_t plane = 0;
+                u32 plane = 0;
                 if (weightParams.m_bDualPlane && (((planeIdx + 1) & 3) == c)) {
                     plane = 1;
                 }
 
-                uint32_t weight = weights[plane][j * blockWidth + i];
-                uint32_t C = (C0 * (64 - weight) + C1 * weight + 32) / 64;
+                u32 weight = weights[plane][j * blockWidth + i];
+                u32 C = (C0 * (64 - weight) + C1 * weight + 32) / 64;
                 if (C == 65535) {
                     p.Component(c) = 255;
                 } else {
                     double Cf = static_cast<double>(C);
-                    p.Component(c) = static_cast<uint16_t>(255.0 * (Cf / 65536.0) + 0.5);
+                    p.Component(c) = static_cast<u16>(255.0 * (Cf / 65536.0) + 0.5);
                 }
             }
 
@@ -1613,26 +1595,26 @@ static void DecompressBlock(const uint8_t inBuf[16], const uint32_t blockWidth,
 
 namespace Tegra::Texture::ASTC {
 
-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<u8> Decompress(const u8* data, u32 width, u32 height, u32 depth, u32 block_width,
+                           u32 block_height) {
+    u32 blockIdx = 0;
     std::size_t depth_offset = 0;
-    std::vector<uint8_t> outData(height * width * depth * 4);
-    for (uint32_t k = 0; k < depth; k++) {
-        for (uint32_t j = 0; j < height; j += block_height) {
-            for (uint32_t i = 0; i < width; i += block_width) {
+    std::vector<u8> outData(height * width * depth * 4);
+    for (u32 k = 0; k < depth; k++) {
+        for (u32 j = 0; j < height; j += block_height) {
+            for (u32 i = 0; i < width; i += block_width) {
 
-                const uint8_t* blockPtr = data + blockIdx * 16;
+                const u8* blockPtr = data + blockIdx * 16;
 
                 // Blocks can be at most 12x12
-                uint32_t uncompData[144];
+                u32 uncompData[144];
                 ASTCC::DecompressBlock(blockPtr, block_width, block_height, uncompData);
 
-                uint32_t decompWidth = std::min(block_width, width - i);
-                uint32_t decompHeight = std::min(block_height, height - j);
+                u32 decompWidth = std::min(block_width, width - i);
+                u32 decompHeight = std::min(block_height, height - j);
 
-                uint8_t* outRow = depth_offset + outData.data() + (j * width + i) * 4;
-                for (uint32_t jj = 0; jj < decompHeight; jj++) {
+                u8* outRow = depth_offset + outData.data() + (j * width + i) * 4;
+                for (u32 jj = 0; jj < decompHeight; jj++) {
                     memcpy(outRow + jj * width * 4, uncompData + jj * block_width, decompWidth * 4);
                 }
 
diff --git a/src/video_core/textures/texture.h b/src/video_core/textures/texture.h
index 8e82c6748..7edc4abe1 100644
--- a/src/video_core/textures/texture.h
+++ b/src/video_core/textures/texture.h
@@ -8,6 +8,7 @@
 #include "common/assert.h"
 #include "common/bit_field.h"
 #include "common/common_types.h"
+#include "core/settings.h"
 
 namespace Tegra::Texture {
 
@@ -294,6 +295,14 @@ enum class TextureMipmapFilter : u32 {
     Linear = 3,
 };
 
+enum class Anisotropy {
+    Default,
+    Filter2x,
+    Filter4x,
+    Filter8x,
+    Filter16x,
+};
+
 struct TSCEntry {
     union {
         struct {
@@ -328,7 +337,22 @@ struct TSCEntry {
     };
 
     float GetMaxAnisotropy() const {
-        return static_cast<float>(1U << max_anisotropy);
+        const u32 min_value = [] {
+            switch (static_cast<Anisotropy>(Settings::values.max_anisotropy)) {
+            default:
+            case Anisotropy::Default:
+                return 1U;
+            case Anisotropy::Filter2x:
+                return 2U;
+            case Anisotropy::Filter4x:
+                return 4U;
+            case Anisotropy::Filter8x:
+                return 8U;
+            case Anisotropy::Filter16x:
+                return 16U;
+            }
+        }();
+        return static_cast<float>(std::max(1U << max_anisotropy, min_value));
     }
 
     float GetMinLod() const {
diff --git a/src/yuzu/CMakeLists.txt b/src/yuzu/CMakeLists.txt
index b841e63fa..d34b47b3f 100644
--- a/src/yuzu/CMakeLists.txt
+++ b/src/yuzu/CMakeLists.txt
@@ -42,6 +42,9 @@ add_executable(yuzu
     configuration/configure_graphics.cpp
     configuration/configure_graphics.h
     configuration/configure_graphics.ui
+    configuration/configure_graphics_advanced.cpp
+    configuration/configure_graphics_advanced.h
+    configuration/configure_graphics_advanced.ui
     configuration/configure_hotkeys.cpp
     configuration/configure_hotkeys.h
     configuration/configure_hotkeys.ui
diff --git a/src/yuzu/bootmanager.cpp b/src/yuzu/bootmanager.cpp
index 55a37fffa..c3dbb1a88 100644
--- a/src/yuzu/bootmanager.cpp
+++ b/src/yuzu/bootmanager.cpp
@@ -9,6 +9,9 @@
 #include <QKeyEvent>
 #include <QMessageBox>
 #include <QOffscreenSurface>
+#include <QOpenGLContext>
+#include <QOpenGLFunctions>
+#include <QOpenGLFunctions_4_3_Core>
 #include <QOpenGLWindow>
 #include <QPainter>
 #include <QScreen>
@@ -23,9 +26,10 @@
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "common/scm_rev.h"
+#include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/frontend/framebuffer_layout.h"
-#include "core/frontend/scope_acquire_window_context.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "core/settings.h"
 #include "input_common/keyboard.h"
 #include "input_common/main.h"
@@ -35,15 +39,27 @@
 #include "yuzu/bootmanager.h"
 #include "yuzu/main.h"
 
-EmuThread::EmuThread(GRenderWindow* render_window) : render_window(render_window) {}
+EmuThread::EmuThread(GRenderWindow& window)
+    : shared_context{window.CreateSharedContext()},
+      context{(Settings::values.use_asynchronous_gpu_emulation && shared_context) ? *shared_context
+                                                                                  : window} {}
 
 EmuThread::~EmuThread() = default;
 
-void EmuThread::run() {
-    render_window->MakeCurrent();
+static GMainWindow* GetMainWindow() {
+    for (QWidget* w : qApp->topLevelWidgets()) {
+        if (GMainWindow* main = qobject_cast<GMainWindow*>(w)) {
+            return main;
+        }
+    }
+    return nullptr;
+}
 
+void EmuThread::run() {
     MicroProfileOnThreadCreate("EmuThread");
 
+    Core::Frontend::ScopeAcquireContext acquire_context{context};
+
     emit LoadProgress(VideoCore::LoadCallbackStage::Prepare, 0, 0);
 
     Core::System::GetInstance().Renderer().Rasterizer().LoadDiskResources(
@@ -53,11 +69,6 @@ void EmuThread::run() {
 
     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,
     // so that the DebugModeLeft signal can be emitted before the
     // next execution step
@@ -98,190 +109,202 @@ void EmuThread::run() {
 #if MICROPROFILE_ENABLED
     MicroProfileOnThreadExit();
 #endif
-
-    render_window->moveContext();
 }
 
 class GGLContext : public Core::Frontend::GraphicsContext {
 public:
-    explicit GGLContext(QOpenGLContext* shared_context) : shared_context{shared_context} {
-        context.setFormat(shared_context->format());
-        context.setShareContext(shared_context);
-        context.create();
+    explicit GGLContext(QOpenGLContext* shared_context)
+        : context(new QOpenGLContext(shared_context->parent())),
+          surface(new QOffscreenSurface(nullptr)) {
+
+        // disable vsync for any shared contexts
+        auto format = shared_context->format();
+        format.setSwapInterval(0);
+
+        context->setShareContext(shared_context);
+        context->setFormat(format);
+        context->create();
+        surface->setParent(shared_context->parent());
+        surface->setFormat(format);
+        surface->create();
     }
 
     void MakeCurrent() override {
-        context.makeCurrent(shared_context->surface());
+        context->makeCurrent(surface);
     }
 
     void DoneCurrent() override {
-        context.doneCurrent();
+        context->doneCurrent();
     }
 
-    void SwapBuffers() override {}
-
 private:
-    QOpenGLContext* shared_context;
-    QOpenGLContext context;
+    QOpenGLContext* context;
+    QOffscreenSurface* surface;
 };
 
-class GWidgetInternal : public QWindow {
+class ChildRenderWindow : public QWindow {
 public:
-    GWidgetInternal(GRenderWindow* parent) : parent(parent) {}
-    virtual ~GWidgetInternal() = default;
+    ChildRenderWindow(QWindow* parent, QWidget* event_handler)
+        : QWindow{parent}, event_handler{event_handler} {}
 
-    void resizeEvent(QResizeEvent* ev) override {
-        parent->OnClientAreaResized(ev->size().width(), ev->size().height());
-        parent->OnFramebufferSizeChanged();
-    }
+    virtual ~ChildRenderWindow() = default;
 
-    void keyPressEvent(QKeyEvent* event) override {
-        InputCommon::GetKeyboard()->PressKey(event->key());
-    }
+    virtual void Present() = 0;
 
-    void keyReleaseEvent(QKeyEvent* event) override {
-        InputCommon::GetKeyboard()->ReleaseKey(event->key());
+protected:
+    bool event(QEvent* event) override {
+        switch (event->type()) {
+        case QEvent::UpdateRequest:
+            Present();
+            return true;
+        case QEvent::MouseButtonPress:
+        case QEvent::MouseButtonRelease:
+        case QEvent::MouseButtonDblClick:
+        case QEvent::MouseMove:
+        case QEvent::KeyPress:
+        case QEvent::KeyRelease:
+        case QEvent::FocusIn:
+        case QEvent::FocusOut:
+        case QEvent::FocusAboutToChange:
+        case QEvent::Enter:
+        case QEvent::Leave:
+        case QEvent::Wheel:
+        case QEvent::TabletMove:
+        case QEvent::TabletPress:
+        case QEvent::TabletRelease:
+        case QEvent::TabletEnterProximity:
+        case QEvent::TabletLeaveProximity:
+        case QEvent::TouchBegin:
+        case QEvent::TouchUpdate:
+        case QEvent::TouchEnd:
+        case QEvent::InputMethodQuery:
+        case QEvent::TouchCancel:
+            return QCoreApplication::sendEvent(event_handler, event);
+        case QEvent::Drop:
+            GetMainWindow()->DropAction(static_cast<QDropEvent*>(event));
+            return true;
+        case QEvent::DragResponse:
+        case QEvent::DragEnter:
+        case QEvent::DragLeave:
+        case QEvent::DragMove:
+            GetMainWindow()->AcceptDropEvent(static_cast<QDropEvent*>(event));
+            return true;
+        default:
+            return QWindow::event(event);
+        }
     }
 
-    void mousePressEvent(QMouseEvent* event) override {
-        if (event->source() == Qt::MouseEventSynthesizedBySystem)
-            return; // touch input is handled in TouchBeginEvent
-
-        const auto pos{event->pos()};
-        if (event->button() == Qt::LeftButton) {
-            const auto [x, y] = parent->ScaleTouch(pos);
-            parent->TouchPressed(x, y);
-        } else if (event->button() == Qt::RightButton) {
-            InputCommon::GetMotionEmu()->BeginTilt(pos.x(), pos.y());
-        }
+    void exposeEvent(QExposeEvent* event) override {
+        QWindow::requestUpdate();
+        QWindow::exposeEvent(event);
     }
 
-    void mouseMoveEvent(QMouseEvent* event) override {
-        if (event->source() == Qt::MouseEventSynthesizedBySystem)
-            return; // touch input is handled in TouchUpdateEvent
+private:
+    QWidget* event_handler{};
+};
 
-        const auto pos{event->pos()};
-        const auto [x, y] = parent->ScaleTouch(pos);
-        parent->TouchMoved(x, y);
-        InputCommon::GetMotionEmu()->Tilt(pos.x(), pos.y());
-    }
+class OpenGLWindow final : public ChildRenderWindow {
+public:
+    OpenGLWindow(QWindow* parent, QWidget* event_handler, QOpenGLContext* shared_context)
+        : ChildRenderWindow{parent, event_handler},
+          context(new QOpenGLContext(shared_context->parent())) {
 
-    void mouseReleaseEvent(QMouseEvent* event) override {
-        if (event->source() == Qt::MouseEventSynthesizedBySystem)
-            return; // touch input is handled in TouchEndEvent
+        // disable vsync for any shared contexts
+        auto format = shared_context->format();
+        format.setSwapInterval(Settings::values.use_vsync ? 1 : 0);
+        this->setFormat(format);
 
-        if (event->button() == Qt::LeftButton)
-            parent->TouchReleased();
-        else if (event->button() == Qt::RightButton)
-            InputCommon::GetMotionEmu()->EndTilt();
-    }
+        context->setShareContext(shared_context);
+        context->setScreen(this->screen());
+        context->setFormat(format);
+        context->create();
 
-    void DisablePainting() {
-        do_painting = false;
-    }
+        setSurfaceType(QWindow::OpenGLSurface);
 
-    void EnablePainting() {
-        do_painting = true;
+        // TODO: One of these flags might be interesting: WA_OpaquePaintEvent, WA_NoBackground,
+        // WA_DontShowOnScreen, WA_DeleteOnClose
     }
 
-    std::pair<unsigned, unsigned> GetSize() const {
-        return std::make_pair(width(), height());
+    ~OpenGLWindow() override {
+        context->doneCurrent();
     }
 
-protected:
-    bool IsPaintingEnabled() const {
-        return do_painting;
+    void Present() override {
+        if (!isExposed()) {
+            return;
+        }
+
+        context->makeCurrent(this);
+        Core::System::GetInstance().Renderer().TryPresent(100);
+        context->swapBuffers(this);
+        auto f = context->versionFunctions<QOpenGLFunctions_4_3_Core>();
+        f->glFinish();
+        QWindow::requestUpdate();
     }
 
 private:
-    GRenderWindow* parent;
-    bool do_painting = false;
-};
-
-// This class overrides paintEvent and resizeEvent to prevent the GUI thread from stealing GL
-// context.
-// The corresponding functionality is handled in EmuThread instead
-class GGLWidgetInternal final : public GWidgetInternal, public QOpenGLWindow {
-public:
-    GGLWidgetInternal(GRenderWindow* parent, QOpenGLContext* shared_context)
-        : GWidgetInternal(parent), QOpenGLWindow(shared_context) {}
-    ~GGLWidgetInternal() override = default;
-
-    void paintEvent(QPaintEvent* ev) override {
-        if (IsPaintingEnabled()) {
-            QPainter painter(this);
-        }
-    }
+    QOpenGLContext* context{};
 };
 
 #ifdef HAS_VULKAN
-class GVKWidgetInternal final : public GWidgetInternal {
+class VulkanWindow final : public ChildRenderWindow {
 public:
-    GVKWidgetInternal(GRenderWindow* parent, QVulkanInstance* instance) : GWidgetInternal(parent) {
+    VulkanWindow(QWindow* parent, QWidget* event_handler, QVulkanInstance* instance)
+        : ChildRenderWindow{parent, event_handler} {
         setSurfaceType(QSurface::SurfaceType::VulkanSurface);
         setVulkanInstance(instance);
     }
-    ~GVKWidgetInternal() override = default;
+
+    ~VulkanWindow() override = default;
+
+    void Present() override {
+        // TODO(bunnei): ImplementMe
+    }
+
+private:
+    QWidget* event_handler{};
 };
 #endif
 
-GRenderWindow::GRenderWindow(GMainWindow* parent, EmuThread* emu_thread)
-    : QWidget(parent), emu_thread(emu_thread) {
+GRenderWindow::GRenderWindow(QWidget* parent_, EmuThread* emu_thread)
+    : QWidget(parent_), emu_thread(emu_thread) {
     setWindowTitle(QStringLiteral("yuzu %1 | %2-%3")
                        .arg(QString::fromUtf8(Common::g_build_name),
                             QString::fromUtf8(Common::g_scm_branch),
                             QString::fromUtf8(Common::g_scm_desc)));
     setAttribute(Qt::WA_AcceptTouchEvents);
-
+    auto layout = new QHBoxLayout(this);
+    layout->setMargin(0);
+    setLayout(layout);
     InputCommon::Init();
+
+    GMainWindow* parent = GetMainWindow();
     connect(this, &GRenderWindow::FirstFrameDisplayed, parent, &GMainWindow::OnLoadComplete);
 }
 
 GRenderWindow::~GRenderWindow() {
     InputCommon::Shutdown();
-
-    // Avoid an unordered destruction that generates a segfault
-    delete child;
 }
 
-void GRenderWindow::moveContext() {
-    if (!context) {
-        return;
+void GRenderWindow::MakeCurrent() {
+    if (core_context) {
+        core_context->MakeCurrent();
     }
-    DoneCurrent();
-
-    // If the thread started running, move the GL Context to the new thread. Otherwise, move it
-    // back.
-    auto thread = (QThread::currentThread() == qApp->thread() && emu_thread != nullptr)
-                      ? emu_thread
-                      : qApp->thread();
-    context->moveToThread(thread);
 }
 
-void GRenderWindow::SwapBuffers() {
-    if (context) {
-        context->swapBuffers(child);
+void GRenderWindow::DoneCurrent() {
+    if (core_context) {
+        core_context->DoneCurrent();
     }
+}
+
+void GRenderWindow::PollEvents() {
     if (!first_frame) {
         first_frame = true;
         emit FirstFrameDisplayed();
     }
 }
 
-void GRenderWindow::MakeCurrent() {
-    if (context) {
-        context->makeCurrent(child);
-    }
-}
-
-void GRenderWindow::DoneCurrent() {
-    if (context) {
-        context->doneCurrent();
-    }
-}
-
-void GRenderWindow::PollEvents() {}
-
 bool GRenderWindow::IsShown() const {
     return !isMinimized();
 }
@@ -291,7 +314,7 @@ void GRenderWindow::RetrieveVulkanHandlers(void* get_instance_proc_addr, void* i
 #ifdef HAS_VULKAN
     const auto instance_proc_addr = vk_instance->getInstanceProcAddr("vkGetInstanceProcAddr");
     const VkInstance instance_copy = vk_instance->vkInstance();
-    const VkSurfaceKHR surface_copy = vk_instance->surfaceForWindow(child);
+    const VkSurfaceKHR surface_copy = vk_instance->surfaceForWindow(child_window);
 
     std::memcpy(get_instance_proc_addr, &instance_proc_addr, sizeof(instance_proc_addr));
     std::memcpy(instance, &instance_copy, sizeof(instance_copy));
@@ -309,21 +332,10 @@ void GRenderWindow::RetrieveVulkanHandlers(void* get_instance_proc_addr, void* i
 void GRenderWindow::OnFramebufferSizeChanged() {
     // Screen changes potentially incur a change in screen DPI, hence we should update the
     // framebuffer size
-    const qreal pixelRatio{GetWindowPixelRatio()};
-    const auto size{child->GetSize()};
-    UpdateCurrentFramebufferLayout(size.first * pixelRatio, size.second * pixelRatio);
-}
-
-void GRenderWindow::ForwardKeyPressEvent(QKeyEvent* event) {
-    if (child) {
-        child->keyPressEvent(event);
-    }
-}
-
-void GRenderWindow::ForwardKeyReleaseEvent(QKeyEvent* event) {
-    if (child) {
-        child->keyReleaseEvent(event);
-    }
+    const qreal pixel_ratio = windowPixelRatio();
+    const u32 width = this->width() * pixel_ratio;
+    const u32 height = this->height() * pixel_ratio;
+    UpdateCurrentFramebufferLayout(width, height);
 }
 
 void GRenderWindow::BackupGeometry() {
@@ -351,13 +363,12 @@ QByteArray GRenderWindow::saveGeometry() {
     return geometry;
 }
 
-qreal GRenderWindow::GetWindowPixelRatio() const {
-    // windowHandle() might not be accessible until the window is displayed to screen.
-    return windowHandle() ? windowHandle()->screen()->devicePixelRatio() : 1.0f;
+qreal GRenderWindow::windowPixelRatio() const {
+    return devicePixelRatio();
 }
 
 std::pair<u32, u32> GRenderWindow::ScaleTouch(const QPointF pos) const {
-    const qreal pixel_ratio{GetWindowPixelRatio()};
+    const qreal pixel_ratio = windowPixelRatio();
     return {static_cast<u32>(std::max(std::round(pos.x() * pixel_ratio), qreal{0.0})),
             static_cast<u32>(std::max(std::round(pos.y() * pixel_ratio), qreal{0.0}))};
 }
@@ -367,6 +378,47 @@ void GRenderWindow::closeEvent(QCloseEvent* event) {
     QWidget::closeEvent(event);
 }
 
+void GRenderWindow::keyPressEvent(QKeyEvent* event) {
+    InputCommon::GetKeyboard()->PressKey(event->key());
+}
+
+void GRenderWindow::keyReleaseEvent(QKeyEvent* event) {
+    InputCommon::GetKeyboard()->ReleaseKey(event->key());
+}
+
+void GRenderWindow::mousePressEvent(QMouseEvent* event) {
+    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        return; // touch input is handled in TouchBeginEvent
+
+    auto pos = event->pos();
+    if (event->button() == Qt::LeftButton) {
+        const auto [x, y] = ScaleTouch(pos);
+        this->TouchPressed(x, y);
+    } else if (event->button() == Qt::RightButton) {
+        InputCommon::GetMotionEmu()->BeginTilt(pos.x(), pos.y());
+    }
+}
+
+void GRenderWindow::mouseMoveEvent(QMouseEvent* event) {
+    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        return; // touch input is handled in TouchUpdateEvent
+
+    auto pos = event->pos();
+    const auto [x, y] = ScaleTouch(pos);
+    this->TouchMoved(x, y);
+    InputCommon::GetMotionEmu()->Tilt(pos.x(), pos.y());
+}
+
+void GRenderWindow::mouseReleaseEvent(QMouseEvent* event) {
+    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        return; // touch input is handled in TouchEndEvent
+
+    if (event->button() == Qt::LeftButton)
+        this->TouchReleased();
+    else if (event->button() == Qt::RightButton)
+        InputCommon::GetMotionEmu()->EndTilt();
+}
+
 void GRenderWindow::TouchBeginEvent(const QTouchEvent* event) {
     // TouchBegin always has exactly one touch point, so take the .first()
     const auto [x, y] = ScaleTouch(event->touchPoints().first().pos());
@@ -415,26 +467,20 @@ void GRenderWindow::focusOutEvent(QFocusEvent* event) {
     InputCommon::GetKeyboard()->ReleaseAllKeys();
 }
 
-void GRenderWindow::OnClientAreaResized(u32 width, u32 height) {
-    NotifyClientAreaSizeChanged(std::make_pair(width, height));
+void GRenderWindow::resizeEvent(QResizeEvent* event) {
+    QWidget::resizeEvent(event);
+    OnFramebufferSizeChanged();
 }
 
 std::unique_ptr<Core::Frontend::GraphicsContext> GRenderWindow::CreateSharedContext() const {
-    return std::make_unique<GGLContext>(context.get());
+    if (Settings::values.renderer_backend == Settings::RendererBackend::OpenGL) {
+        return std::make_unique<GGLContext>(QOpenGLContext::globalShareContext());
+    }
+    return {};
 }
 
 bool GRenderWindow::InitRenderTarget() {
-    shared_context.reset();
-    context.reset();
-    if (child) {
-        delete child;
-    }
-    if (container) {
-        delete container;
-    }
-    if (layout()) {
-        delete layout();
-    }
+    ReleaseRenderTarget();
 
     first_frame = false;
 
@@ -451,13 +497,6 @@ bool GRenderWindow::InitRenderTarget() {
         break;
     }
 
-    container = QWidget::createWindowContainer(child, this);
-    QBoxLayout* layout = new QHBoxLayout(this);
-
-    layout->addWidget(container);
-    layout->setMargin(0);
-    setLayout(layout);
-
     // Reset minimum required size to avoid resizing issues on the main window after restarting.
     setMinimumSize(1, 1);
 
@@ -467,14 +506,9 @@ bool GRenderWindow::InitRenderTarget() {
     hide();
 
     resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
-    child->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
-    container->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
 
     OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
-
     OnFramebufferSizeChanged();
-    NotifyClientAreaSizeChanged(child->GetSize());
-
     BackupGeometry();
 
     if (Settings::values.renderer_backend == Settings::RendererBackend::OpenGL) {
@@ -486,6 +520,14 @@ bool GRenderWindow::InitRenderTarget() {
     return true;
 }
 
+void GRenderWindow::ReleaseRenderTarget() {
+    if (child_widget) {
+        layout()->removeWidget(child_widget);
+        delete child_widget;
+        child_widget = nullptr;
+    }
+}
+
 void GRenderWindow::CaptureScreenshot(u32 res_scale, const QString& screenshot_path) {
     auto& renderer = Core::System::GetInstance().Renderer();
 
@@ -521,16 +563,19 @@ bool GRenderWindow::InitializeOpenGL() {
     fmt.setOption(QSurfaceFormat::FormatOption::DeprecatedFunctions);
     // TODO: expose a setting for buffer value (ie default/single/double/triple)
     fmt.setSwapBehavior(QSurfaceFormat::DefaultSwapBehavior);
-    shared_context = std::make_unique<QOpenGLContext>();
-    shared_context->setFormat(fmt);
-    shared_context->create();
-    context = std::make_unique<QOpenGLContext>();
-    context->setShareContext(shared_context.get());
-    context->setFormat(fmt);
-    context->create();
-    fmt.setSwapInterval(false);
-
-    child = new GGLWidgetInternal(this, shared_context.get());
+    fmt.setSwapInterval(0);
+    QSurfaceFormat::setDefaultFormat(fmt);
+
+    GMainWindow* parent = GetMainWindow();
+    QWindow* parent_win_handle = parent ? parent->windowHandle() : nullptr;
+    child_window = new OpenGLWindow(parent_win_handle, this, QOpenGLContext::globalShareContext());
+    child_window->create();
+    child_widget = createWindowContainer(child_window, this);
+    child_widget->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
+    layout()->addWidget(child_widget);
+
+    core_context = CreateSharedContext();
+
     return true;
 }
 
@@ -559,7 +604,14 @@ bool GRenderWindow::InitializeVulkan() {
         return false;
     }
 
-    child = new GVKWidgetInternal(this, vk_instance.get());
+    GMainWindow* parent = GetMainWindow();
+    QWindow* parent_win_handle = parent ? parent->windowHandle() : nullptr;
+    child_window = new VulkanWindow(parent_win_handle, this, vk_instance.get());
+    child_window->create();
+    child_widget = createWindowContainer(child_window, this);
+    child_widget->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
+    layout()->addWidget(child_widget);
+
     return true;
 #else
     QMessageBox::critical(this, tr("Vulkan not available!"),
@@ -569,7 +621,7 @@ bool GRenderWindow::InitializeVulkan() {
 }
 
 bool GRenderWindow::LoadOpenGL() {
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{*this};
+    Core::Frontend::ScopeAcquireContext acquire_context{*this};
     if (!gladLoadGL()) {
         QMessageBox::critical(this, tr("Error while initializing OpenGL 4.3!"),
                               tr("Your GPU may not support OpenGL 4.3, or you do not have the "
@@ -621,12 +673,10 @@ QStringList GRenderWindow::GetUnsupportedGLExtensions() const {
 
 void GRenderWindow::OnEmulationStarting(EmuThread* emu_thread) {
     this->emu_thread = emu_thread;
-    child->DisablePainting();
 }
 
 void GRenderWindow::OnEmulationStopping() {
     emu_thread = nullptr;
-    child->EnablePainting();
 }
 
 void GRenderWindow::showEvent(QShowEvent* event) {
diff --git a/src/yuzu/bootmanager.h b/src/yuzu/bootmanager.h
index 71a2fa321..79b030304 100644
--- a/src/yuzu/bootmanager.h
+++ b/src/yuzu/bootmanager.h
@@ -11,11 +11,13 @@
 #include <QImage>
 #include <QThread>
 #include <QWidget>
+#include <QWindow>
 
 #include "common/thread.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
 
+class GRenderWindow;
 class QKeyEvent;
 class QScreen;
 class QTouchEvent;
@@ -26,14 +28,6 @@ class QOpenGLContext;
 class QVulkanInstance;
 #endif
 
-class GWidgetInternal;
-class GGLWidgetInternal;
-class GVKWidgetInternal;
-class GMainWindow;
-class GRenderWindow;
-class QSurface;
-class QOpenGLContext;
-
 namespace VideoCore {
 enum class LoadCallbackStage;
 }
@@ -42,7 +36,7 @@ class EmuThread final : public QThread {
     Q_OBJECT
 
 public:
-    explicit EmuThread(GRenderWindow* render_window);
+    explicit EmuThread(GRenderWindow& window);
     ~EmuThread() override;
 
     /**
@@ -96,7 +90,11 @@ private:
     std::mutex running_mutex;
     std::condition_variable running_cv;
 
-    GRenderWindow* render_window;
+    /// Only used in asynchronous GPU mode
+    std::unique_ptr<Core::Frontend::GraphicsContext> shared_context;
+
+    /// This is shared_context in asynchronous GPU mode, core_context in synchronous GPU mode
+    Core::Frontend::GraphicsContext& context;
 
 signals:
     /**
@@ -126,11 +124,10 @@ class GRenderWindow : public QWidget, public Core::Frontend::EmuWindow {
     Q_OBJECT
 
 public:
-    GRenderWindow(GMainWindow* parent, EmuThread* emu_thread);
+    GRenderWindow(QWidget* parent, EmuThread* emu_thread);
     ~GRenderWindow() override;
 
-    // EmuWindow implementation
-    void SwapBuffers() override;
+    // EmuWindow implementation.
     void MakeCurrent() override;
     void DoneCurrent() override;
     void PollEvents() override;
@@ -139,30 +136,36 @@ public:
                                 void* surface) const override;
     std::unique_ptr<Core::Frontend::GraphicsContext> CreateSharedContext() const override;
 
-    void ForwardKeyPressEvent(QKeyEvent* event);
-    void ForwardKeyReleaseEvent(QKeyEvent* event);
-
     void BackupGeometry();
     void RestoreGeometry();
     void restoreGeometry(const QByteArray& geometry); // overridden
     QByteArray saveGeometry();                        // overridden
 
-    qreal GetWindowPixelRatio() const;
-    std::pair<u32, u32> ScaleTouch(QPointF pos) const;
+    qreal windowPixelRatio() const;
 
     void closeEvent(QCloseEvent* event) override;
+
+    void resizeEvent(QResizeEvent* event) override;
+
+    void keyPressEvent(QKeyEvent* event) override;
+    void keyReleaseEvent(QKeyEvent* event) override;
+
+    void mousePressEvent(QMouseEvent* event) override;
+    void mouseMoveEvent(QMouseEvent* event) override;
+    void mouseReleaseEvent(QMouseEvent* event) override;
+
     bool event(QEvent* event) override;
-    void focusOutEvent(QFocusEvent* event) override;
 
-    void OnClientAreaResized(u32 width, u32 height);
+    void focusOutEvent(QFocusEvent* event) override;
 
     bool InitRenderTarget();
 
+    /// Destroy the previous run's child_widget which should also destroy the child_window
+    void ReleaseRenderTarget();
+
     void CaptureScreenshot(u32 res_scale, const QString& screenshot_path);
 
 public slots:
-    void moveContext(); // overridden
-
     void OnEmulationStarting(EmuThread* emu_thread);
     void OnEmulationStopping();
     void OnFramebufferSizeChanged();
@@ -173,6 +176,7 @@ signals:
     void FirstFrameDisplayed();
 
 private:
+    std::pair<u32, u32> ScaleTouch(QPointF pos) const;
     void TouchBeginEvent(const QTouchEvent* event);
     void TouchUpdateEvent(const QTouchEvent* event);
     void TouchEndEvent();
@@ -184,15 +188,9 @@ private:
     bool LoadOpenGL();
     QStringList GetUnsupportedGLExtensions() const;
 
-    QWidget* container = nullptr;
-    GWidgetInternal* child = nullptr;
-
     EmuThread* emu_thread;
-    // Context that backs the GGLWidgetInternal (and will be used by core to render)
-    std::unique_ptr<QOpenGLContext> context;
-    // Context that will be shared between all newly created contexts. This should never be made
-    // current
-    std::unique_ptr<QOpenGLContext> shared_context;
+
+    std::unique_ptr<GraphicsContext> core_context;
 
 #ifdef HAS_VULKAN
     std::unique_ptr<QVulkanInstance> vk_instance;
@@ -202,6 +200,15 @@ private:
     QImage screenshot_image;
 
     QByteArray geometry;
+
+    /// Native window handle that backs this presentation widget
+    QWindow* child_window = nullptr;
+
+    /// In order to embed the window into GRenderWindow, you need to use createWindowContainer to
+    /// put the child_window into a widget then add it to the layout. This child_widget can be
+    /// parented to GRenderWindow and use Qt's lifetime system
+    QWidget* child_widget = nullptr;
+
     bool first_frame = false;
 
 protected:
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index 6209fff75..3b9ab38dd 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -539,7 +539,7 @@ void Config::ReadDebuggingValues() {
 void Config::ReadServiceValues() {
     qt_config->beginGroup(QStringLiteral("Services"));
     Settings::values.bcat_backend =
-        ReadSetting(QStringLiteral("bcat_backend"), QStringLiteral("boxcat"))
+        ReadSetting(QStringLiteral("bcat_backend"), QStringLiteral("null"))
             .toString()
             .toStdString();
     Settings::values.bcat_boxcat_local =
@@ -631,6 +631,7 @@ void Config::ReadRendererValues() {
     Settings::values.resolution_factor =
         ReadSetting(QStringLiteral("resolution_factor"), 1.0).toFloat();
     Settings::values.aspect_ratio = ReadSetting(QStringLiteral("aspect_ratio"), 0).toInt();
+    Settings::values.max_anisotropy = ReadSetting(QStringLiteral("max_anisotropy"), 0).toInt();
     Settings::values.use_frame_limit =
         ReadSetting(QStringLiteral("use_frame_limit"), true).toBool();
     Settings::values.frame_limit = ReadSetting(QStringLiteral("frame_limit"), 100).toInt();
@@ -640,6 +641,7 @@ void Config::ReadRendererValues() {
         ReadSetting(QStringLiteral("use_accurate_gpu_emulation"), false).toBool();
     Settings::values.use_asynchronous_gpu_emulation =
         ReadSetting(QStringLiteral("use_asynchronous_gpu_emulation"), false).toBool();
+    Settings::values.use_vsync = ReadSetting(QStringLiteral("use_vsync"), true).toBool();
     Settings::values.force_30fps_mode =
         ReadSetting(QStringLiteral("force_30fps_mode"), false).toBool();
 
@@ -680,6 +682,8 @@ void Config::ReadSystemValues() {
 
     Settings::values.language_index = ReadSetting(QStringLiteral("language_index"), 1).toInt();
 
+    Settings::values.region_index = ReadSetting(QStringLiteral("region_index"), 1).toInt();
+
     const auto rng_seed_enabled = ReadSetting(QStringLiteral("rng_seed_enabled"), false).toBool();
     if (rng_seed_enabled) {
         Settings::values.rng_seed = ReadSetting(QStringLiteral("rng_seed"), 0).toULongLong();
@@ -696,6 +700,8 @@ void Config::ReadSystemValues() {
         Settings::values.custom_rtc = std::nullopt;
     }
 
+    Settings::values.sound_index = ReadSetting(QStringLiteral("sound_index"), 1).toInt();
+
     qt_config->endGroup();
 }
 
@@ -1066,6 +1072,7 @@ void Config::SaveRendererValues() {
     WriteSetting(QStringLiteral("resolution_factor"),
                  static_cast<double>(Settings::values.resolution_factor), 1.0);
     WriteSetting(QStringLiteral("aspect_ratio"), Settings::values.aspect_ratio, 0);
+    WriteSetting(QStringLiteral("max_anisotropy"), Settings::values.max_anisotropy, 0);
     WriteSetting(QStringLiteral("use_frame_limit"), Settings::values.use_frame_limit, true);
     WriteSetting(QStringLiteral("frame_limit"), Settings::values.frame_limit, 100);
     WriteSetting(QStringLiteral("use_disk_shader_cache"), Settings::values.use_disk_shader_cache,
@@ -1074,6 +1081,7 @@ void Config::SaveRendererValues() {
                  Settings::values.use_accurate_gpu_emulation, false);
     WriteSetting(QStringLiteral("use_asynchronous_gpu_emulation"),
                  Settings::values.use_asynchronous_gpu_emulation, false);
+    WriteSetting(QStringLiteral("use_vsync"), Settings::values.use_vsync, true);
     WriteSetting(QStringLiteral("force_30fps_mode"), Settings::values.force_30fps_mode, false);
 
     // Cast to double because Qt's written float values are not human-readable
@@ -1110,6 +1118,7 @@ void Config::SaveSystemValues() {
     WriteSetting(QStringLiteral("use_docked_mode"), Settings::values.use_docked_mode, false);
     WriteSetting(QStringLiteral("current_user"), Settings::values.current_user, 0);
     WriteSetting(QStringLiteral("language_index"), Settings::values.language_index, 1);
+    WriteSetting(QStringLiteral("region_index"), Settings::values.region_index, 1);
 
     WriteSetting(QStringLiteral("rng_seed_enabled"), Settings::values.rng_seed.has_value(), false);
     WriteSetting(QStringLiteral("rng_seed"), Settings::values.rng_seed.value_or(0), 0);
@@ -1121,6 +1130,8 @@ void Config::SaveSystemValues() {
                      Settings::values.custom_rtc.value_or(std::chrono::seconds{}).count()),
                  0);
 
+    WriteSetting(QStringLiteral("sound_index"), Settings::values.sound_index, 1);
+
     qt_config->endGroup();
 }
 
diff --git a/src/yuzu/configuration/configure.ui b/src/yuzu/configuration/configure.ui
index 67b990f1a..9aec1bd09 100644
--- a/src/yuzu/configuration/configure.ui
+++ b/src/yuzu/configuration/configure.ui
@@ -83,6 +83,11 @@
          <string>Graphics</string>
         </attribute>
        </widget>
+       <widget class="ConfigureGraphicsAdvanced" name="graphicsAdvancedTab">
+        <attribute name="title">
+         <string>GraphicsAdvanced</string>
+        </attribute>
+       </widget>
        <widget class="ConfigureAudio" name="audioTab">
         <attribute name="title">
          <string>Audio</string>
@@ -160,6 +165,12 @@
    <container>1</container>
   </customwidget>
   <customwidget>
+   <class>ConfigureGraphicsAdvanced</class>
+   <extends>QWidget</extends>
+   <header>configuration/configure_graphics_advanced.h</header>
+   <container>1</container>
+  </customwidget>
+  <customwidget>
    <class>ConfigureWeb</class>
    <extends>QWidget</extends>
    <header>configuration/configure_web.h</header>
diff --git a/src/yuzu/configuration/configure_dialog.cpp b/src/yuzu/configuration/configure_dialog.cpp
index db3b19352..df4473b46 100644
--- a/src/yuzu/configuration/configure_dialog.cpp
+++ b/src/yuzu/configuration/configure_dialog.cpp
@@ -41,6 +41,7 @@ void ConfigureDialog::ApplyConfiguration() {
     ui->inputTab->ApplyConfiguration();
     ui->hotkeysTab->ApplyConfiguration(registry);
     ui->graphicsTab->ApplyConfiguration();
+    ui->graphicsAdvancedTab->ApplyConfiguration();
     ui->audioTab->ApplyConfiguration();
     ui->debugTab->ApplyConfiguration();
     ui->webTab->ApplyConfiguration();
@@ -76,7 +77,7 @@ void ConfigureDialog::PopulateSelectionList() {
     const std::array<std::pair<QString, QList<QWidget*>>, 5> items{
         {{tr("General"), {ui->generalTab, ui->webTab, ui->debugTab, ui->uiTab}},
          {tr("System"), {ui->systemTab, ui->profileManagerTab, ui->serviceTab, ui->filesystemTab}},
-         {tr("Graphics"), {ui->graphicsTab}},
+         {tr("Graphics"), {ui->graphicsTab, ui->graphicsAdvancedTab}},
          {tr("Audio"), {ui->audioTab}},
          {tr("Controls"), {ui->inputTab, ui->hotkeysTab}}},
     };
@@ -105,6 +106,7 @@ void ConfigureDialog::UpdateVisibleTabs() {
         {ui->inputTab, tr("Input")},
         {ui->hotkeysTab, tr("Hotkeys")},
         {ui->graphicsTab, tr("Graphics")},
+        {ui->graphicsAdvancedTab, tr("Advanced")},
         {ui->audioTab, tr("Audio")},
         {ui->debugTab, tr("Debug")},
         {ui->webTab, tr("Web")},
diff --git a/src/yuzu/configuration/configure_graphics.cpp b/src/yuzu/configuration/configure_graphics.cpp
index ea899c080..a821c7b3c 100644
--- a/src/yuzu/configuration/configure_graphics.cpp
+++ b/src/yuzu/configuration/configure_graphics.cpp
@@ -100,11 +100,8 @@ void ConfigureGraphics::SetConfiguration() {
     ui->aspect_ratio_combobox->setCurrentIndex(Settings::values.aspect_ratio);
     ui->use_disk_shader_cache->setEnabled(runtime_lock);
     ui->use_disk_shader_cache->setChecked(Settings::values.use_disk_shader_cache);
-    ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
     ui->use_asynchronous_gpu_emulation->setEnabled(runtime_lock);
     ui->use_asynchronous_gpu_emulation->setChecked(Settings::values.use_asynchronous_gpu_emulation);
-    ui->force_30fps_mode->setEnabled(runtime_lock);
-    ui->force_30fps_mode->setChecked(Settings::values.force_30fps_mode);
     UpdateBackgroundColorButton(QColor::fromRgbF(Settings::values.bg_red, Settings::values.bg_green,
                                                  Settings::values.bg_blue));
     UpdateDeviceComboBox();
@@ -117,10 +114,8 @@ void ConfigureGraphics::ApplyConfiguration() {
         ToResolutionFactor(static_cast<Resolution>(ui->resolution_factor_combobox->currentIndex()));
     Settings::values.aspect_ratio = ui->aspect_ratio_combobox->currentIndex();
     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.force_30fps_mode = ui->force_30fps_mode->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());
diff --git a/src/yuzu/configuration/configure_graphics.ui b/src/yuzu/configuration/configure_graphics.ui
index db60426ab..c816d6108 100644
--- a/src/yuzu/configuration/configure_graphics.ui
+++ b/src/yuzu/configuration/configure_graphics.ui
@@ -85,20 +85,6 @@
          </widget>
         </item>
         <item>
-         <widget class="QCheckBox" name="use_accurate_gpu_emulation">
-          <property name="text">
-           <string>Use accurate GPU emulation (slow)</string>
-          </property>
-         </widget>
-        </item>
-        <item>
-         <widget class="QCheckBox" name="force_30fps_mode">
-          <property name="text">
-           <string>Force 30 FPS mode</string>
-          </property>
-         </widget>
-        </item>
-        <item>
          <layout class="QHBoxLayout" name="horizontalLayout_2">
           <item>
            <widget class="QLabel" name="label">
diff --git a/src/yuzu/configuration/configure_graphics_advanced.cpp b/src/yuzu/configuration/configure_graphics_advanced.cpp
new file mode 100644
index 000000000..b9f429f84
--- /dev/null
+++ b/src/yuzu/configuration/configure_graphics_advanced.cpp
@@ -0,0 +1,48 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "core/core.h"
+#include "core/settings.h"
+#include "ui_configure_graphics_advanced.h"
+#include "yuzu/configuration/configure_graphics_advanced.h"
+
+ConfigureGraphicsAdvanced::ConfigureGraphicsAdvanced(QWidget* parent)
+    : QWidget(parent), ui(new Ui::ConfigureGraphicsAdvanced) {
+
+    ui->setupUi(this);
+
+    SetConfiguration();
+}
+
+ConfigureGraphicsAdvanced::~ConfigureGraphicsAdvanced() = default;
+
+void ConfigureGraphicsAdvanced::SetConfiguration() {
+    const bool runtime_lock = !Core::System::GetInstance().IsPoweredOn();
+    ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
+    ui->use_vsync->setEnabled(runtime_lock);
+    ui->use_vsync->setChecked(Settings::values.use_vsync);
+    ui->force_30fps_mode->setEnabled(runtime_lock);
+    ui->force_30fps_mode->setChecked(Settings::values.force_30fps_mode);
+    ui->anisotropic_filtering_combobox->setEnabled(runtime_lock);
+    ui->anisotropic_filtering_combobox->setCurrentIndex(Settings::values.max_anisotropy);
+}
+
+void ConfigureGraphicsAdvanced::ApplyConfiguration() {
+    Settings::values.use_accurate_gpu_emulation = ui->use_accurate_gpu_emulation->isChecked();
+    Settings::values.use_vsync = ui->use_vsync->isChecked();
+    Settings::values.force_30fps_mode = ui->force_30fps_mode->isChecked();
+    Settings::values.max_anisotropy = ui->anisotropic_filtering_combobox->currentIndex();
+}
+
+void ConfigureGraphicsAdvanced::changeEvent(QEvent* event) {
+    if (event->type() == QEvent::LanguageChange) {
+        RetranslateUI();
+    }
+
+    QWidget::changeEvent(event);
+}
+
+void ConfigureGraphicsAdvanced::RetranslateUI() {
+    ui->retranslateUi(this);
+}
diff --git a/src/yuzu/configuration/configure_graphics_advanced.h b/src/yuzu/configuration/configure_graphics_advanced.h
new file mode 100644
index 000000000..bbc9d4355
--- /dev/null
+++ b/src/yuzu/configuration/configure_graphics_advanced.h
@@ -0,0 +1,30 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <QWidget>
+
+namespace Ui {
+class ConfigureGraphicsAdvanced;
+}
+
+class ConfigureGraphicsAdvanced : public QWidget {
+    Q_OBJECT
+
+public:
+    explicit ConfigureGraphicsAdvanced(QWidget* parent = nullptr);
+    ~ConfigureGraphicsAdvanced() override;
+
+    void ApplyConfiguration();
+
+private:
+    void changeEvent(QEvent* event) override;
+    void RetranslateUI();
+
+    void SetConfiguration();
+
+    std::unique_ptr<Ui::ConfigureGraphicsAdvanced> ui;
+};
diff --git a/src/yuzu/configuration/configure_graphics_advanced.ui b/src/yuzu/configuration/configure_graphics_advanced.ui
new file mode 100644
index 000000000..42eec278e
--- /dev/null
+++ b/src/yuzu/configuration/configure_graphics_advanced.ui
@@ -0,0 +1,111 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<ui version="4.0">
+ <class>ConfigureGraphicsAdvanced</class>
+ <widget class="QWidget" name="ConfigureGraphicsAdvanced">
+  <property name="geometry">
+   <rect>
+    <x>0</x>
+    <y>0</y>
+    <width>400</width>
+    <height>321</height>
+   </rect>
+  </property>
+  <property name="windowTitle">
+   <string>Form</string>
+  </property>
+  <layout class="QVBoxLayout" name="verticalLayout_1">
+   <item>
+    <layout class="QVBoxLayout" name="verticalLayout_2">
+     <item>
+      <widget class="QGroupBox" name="groupBox_1">
+       <property name="title">
+        <string>Advanced Graphics Settings</string>
+       </property>
+       <layout class="QVBoxLayout" name="verticalLayout_3">
+        <item>
+         <widget class="QCheckBox" name="use_accurate_gpu_emulation">
+          <property name="text">
+           <string>Use accurate GPU emulation (slow)</string>
+          </property>
+         </widget>
+        </item>
+        <item>
+         <widget class="QCheckBox" name="use_vsync">
+          <property name="toolTip">
+           <string>VSync prevents the screen from tearing, but some graphics cards have lower performance with VSync enabled. Keep it enabled if you don't notice a performance difference.</string>
+          </property>
+          <property name="text">
+           <string>Use VSync (OpenGL only)</string>
+          </property>
+         </widget>
+        </item>
+        <item>
+         <widget class="QCheckBox" name="force_30fps_mode">
+          <property name="text">
+           <string>Force 30 FPS mode</string>
+          </property>
+         </widget>
+        </item>
+        <item>
+         <layout class="QHBoxLayout" name="horizontalLayout_1">
+          <item>
+           <widget class="QLabel" name="af_label">
+            <property name="text">
+             <string>Anisotropic Filtering:</string>
+            </property>
+           </widget>
+          </item>
+          <item>
+           <widget class="QComboBox" name="anisotropic_filtering_combobox">
+            <item>
+             <property name="text">
+              <string>Default</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string>2x</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string>4x</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string>8x</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string>16x</string>
+             </property>
+            </item>
+           </widget>
+          </item>
+         </layout>
+        </item>
+       </layout>
+      </widget>
+     </item>
+    </layout>
+   </item>
+   <item>
+    <spacer name="verticalSpacer">
+     <property name="orientation">
+      <enum>Qt::Vertical</enum>
+     </property>
+     <property name="sizeHint" stdset="0">
+      <size>
+       <width>20</width>
+       <height>40</height>
+      </size>
+     </property>
+    </spacer>
+   </item>
+  </layout>
+ </widget>
+ <resources/>
+ <connections/>
+</ui>
diff --git a/src/yuzu/configuration/configure_system.cpp b/src/yuzu/configuration/configure_system.cpp
index e1b52f8d9..f49cd4c8f 100644
--- a/src/yuzu/configuration/configure_system.cpp
+++ b/src/yuzu/configuration/configure_system.cpp
@@ -56,6 +56,8 @@ void ConfigureSystem::SetConfiguration() {
     enabled = !Core::System::GetInstance().IsPoweredOn();
 
     ui->combo_language->setCurrentIndex(Settings::values.language_index);
+    ui->combo_region->setCurrentIndex(Settings::values.region_index);
+    ui->combo_sound->setCurrentIndex(Settings::values.sound_index);
 
     ui->rng_seed_checkbox->setChecked(Settings::values.rng_seed.has_value());
     ui->rng_seed_edit->setEnabled(Settings::values.rng_seed.has_value());
@@ -81,6 +83,8 @@ void ConfigureSystem::ApplyConfiguration() {
     }
 
     Settings::values.language_index = ui->combo_language->currentIndex();
+    Settings::values.region_index = ui->combo_region->currentIndex();
+    Settings::values.sound_index = ui->combo_sound->currentIndex();
 
     if (ui->rng_seed_checkbox->isChecked()) {
         Settings::values.rng_seed = ui->rng_seed_edit->text().toULongLong(nullptr, 16);
diff --git a/src/yuzu/configuration/configure_system.h b/src/yuzu/configuration/configure_system.h
index 1eab3781d..d8fa2d2cc 100644
--- a/src/yuzu/configuration/configure_system.h
+++ b/src/yuzu/configuration/configure_system.h
@@ -36,5 +36,6 @@ private:
     bool enabled = false;
 
     int language_index = 0;
+    int region_index = 0;
     int sound_index = 0;
 };
diff --git a/src/yuzu/configuration/configure_system.ui b/src/yuzu/configuration/configure_system.ui
index 65745a2f8..4e2c7e76e 100644
--- a/src/yuzu/configuration/configure_system.ui
+++ b/src/yuzu/configuration/configure_system.ui
@@ -22,14 +22,14 @@
         <string>System Settings</string>
        </property>
        <layout class="QGridLayout" name="gridLayout">
-        <item row="1" column="0">
+        <item row="2" column="0">
          <widget class="QLabel" name="label_sound">
           <property name="text">
            <string>Sound output mode</string>
           </property>
          </widget>
         </item>
-        <item row="2" column="0">
+        <item row="3" column="0">
          <widget class="QLabel" name="label_console_id">
           <property name="text">
            <string>Console ID:</string>
@@ -128,14 +128,60 @@
           </item>
          </widget>
         </item>
-        <item row="4" column="0">
+        <item row="1" column="0">
+         <widget class="QLabel" name="label_region">
+          <property name="text">
+           <string>Region:</string>
+          </property>
+         </widget>
+        </item>
+        <item row="1" column="1">
+         <widget class="QComboBox" name="combo_region">
+          <item>
+           <property name="text">
+            <string>Japan</string>
+           </property>
+          </item>
+          <item>
+           <property name="text">
+            <string>USA</string>
+           </property>
+          </item>
+          <item>
+           <property name="text">
+            <string>Europe</string>
+           </property>
+          </item>
+          <item>
+           <property name="text">
+            <string>Australia</string>
+           </property>
+          </item>
+          <item>
+           <property name="text">
+            <string>China</string>
+           </property>
+          </item>
+          <item>
+           <property name="text">
+            <string>Korea</string>
+           </property>
+          </item>
+          <item>
+           <property name="text">
+            <string>Taiwan</string>
+           </property>
+          </item>
+         </widget>
+        </item>
+        <item row="5" column="0">
          <widget class="QCheckBox" name="rng_seed_checkbox">
           <property name="text">
            <string>RNG Seed</string>
           </property>
          </widget>
         </item>
-        <item row="1" column="1">
+        <item row="2" column="1">
          <widget class="QComboBox" name="combo_sound">
           <item>
            <property name="text">
@@ -161,7 +207,7 @@
           </property>
          </widget>
         </item>
-        <item row="2" column="1">
+        <item row="3" column="1">
          <widget class="QPushButton" name="button_regenerate_console_id">
           <property name="sizePolicy">
            <sizepolicy hsizetype="Fixed" vsizetype="Fixed">
@@ -177,14 +223,14 @@
           </property>
          </widget>
         </item>
-        <item row="3" column="0">
+        <item row="4" column="0">
          <widget class="QCheckBox" name="custom_rtc_checkbox">
           <property name="text">
            <string>Custom RTC</string>
           </property>
          </widget>
         </item>
-        <item row="3" column="1">
+        <item row="4" column="1">
          <widget class="QDateTimeEdit" name="custom_rtc_edit">
           <property name="minimumDate">
            <date>
@@ -198,7 +244,7 @@
           </property>
          </widget>
         </item>
-        <item row="4" column="1">
+        <item row="5" column="1">
          <widget class="QLineEdit" name="rng_seed_edit">
           <property name="sizePolicy">
            <sizepolicy hsizetype="Minimum" vsizetype="Fixed">
diff --git a/src/yuzu/debugger/wait_tree.cpp b/src/yuzu/debugger/wait_tree.cpp
index 3f1a94627..c1ea25fb8 100644
--- a/src/yuzu/debugger/wait_tree.cpp
+++ b/src/yuzu/debugger/wait_tree.cpp
@@ -116,7 +116,7 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeCallstack::GetChildren() cons
 
     constexpr std::size_t BaseRegister = 29;
     auto& memory = Core::System::GetInstance().Memory();
-    u64 base_pointer = thread.GetContext().cpu_registers[BaseRegister];
+    u64 base_pointer = thread.GetContext64().cpu_registers[BaseRegister];
 
     while (base_pointer != 0) {
         const u64 lr = memory.Read64(base_pointer + sizeof(u64));
@@ -240,7 +240,7 @@ QString WaitTreeThread::GetText() const {
         break;
     }
 
-    const auto& context = thread.GetContext();
+    const auto& context = thread.GetContext64();
     const QString pc_info = tr(" PC = 0x%1 LR = 0x%2")
                                 .arg(context.pc, 8, 16, QLatin1Char{'0'})
                                 .arg(context.cpu_registers[30], 8, 16, QLatin1Char{'0'});
diff --git a/src/yuzu/loading_screen.cpp b/src/yuzu/loading_screen.cpp
index 4f2bfab48..2a6483370 100644
--- a/src/yuzu/loading_screen.cpp
+++ b/src/yuzu/loading_screen.cpp
@@ -34,18 +34,6 @@ constexpr char PROGRESSBAR_STYLE_PREPARE[] = R"(
 QProgressBar {}
 QProgressBar::chunk {})";
 
-constexpr char PROGRESSBAR_STYLE_DECOMPILE[] = R"(
-QProgressBar {
-  background-color: black;
-  border: 2px solid white;
-  border-radius: 4px;
-  padding: 2px;
-}
-QProgressBar::chunk {
-  background-color: #0ab9e6;
-  width: 1px;
-})";
-
 constexpr char PROGRESSBAR_STYLE_BUILD[] = R"(
 QProgressBar {
   background-color: black;
@@ -100,13 +88,11 @@ LoadingScreen::LoadingScreen(QWidget* parent)
 
     stage_translations = {
         {VideoCore::LoadCallbackStage::Prepare, tr("Loading...")},
-        {VideoCore::LoadCallbackStage::Decompile, tr("Preparing Shaders %1 / %2")},
         {VideoCore::LoadCallbackStage::Build, tr("Loading Shaders %1 / %2")},
         {VideoCore::LoadCallbackStage::Complete, tr("Launching...")},
     };
     progressbar_style = {
         {VideoCore::LoadCallbackStage::Prepare, PROGRESSBAR_STYLE_PREPARE},
-        {VideoCore::LoadCallbackStage::Decompile, PROGRESSBAR_STYLE_DECOMPILE},
         {VideoCore::LoadCallbackStage::Build, PROGRESSBAR_STYLE_BUILD},
         {VideoCore::LoadCallbackStage::Complete, PROGRESSBAR_STYLE_COMPLETE},
     };
@@ -192,8 +178,7 @@ void LoadingScreen::OnLoadProgress(VideoCore::LoadCallbackStage stage, std::size
     }
 
     // update labels and progress bar
-    if (stage == VideoCore::LoadCallbackStage::Decompile ||
-        stage == VideoCore::LoadCallbackStage::Build) {
+    if (stage == VideoCore::LoadCallbackStage::Build) {
         ui->stage->setText(stage_translations[stage].arg(value).arg(total));
     } else {
         ui->stage->setText(stage_translations[stage]);
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index 1be61bd48..4769a612e 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -20,7 +20,6 @@
 #include "core/file_sys/vfs.h"
 #include "core/file_sys/vfs_real.h"
 #include "core/frontend/applets/general_frontend.h"
-#include "core/frontend/scope_acquire_window_context.h"
 #include "core/hle/service/acc/profile_manager.h"
 #include "core/hle/service/am/applet_ae.h"
 #include "core/hle/service/am/applet_oe.h"
@@ -985,11 +984,8 @@ void GMainWindow::BootGame(const QString& filename) {
         return;
 
     // Create and start the emulation thread
-    emu_thread = std::make_unique<EmuThread>(render_window);
+    emu_thread = std::make_unique<EmuThread>(*render_window);
     emit EmulationStarting(emu_thread.get());
-    if (Settings::values.renderer_backend == Settings::RendererBackend::OpenGL) {
-        render_window->moveContext();
-    }
     emu_thread->start();
 
     connect(render_window, &GRenderWindow::Closed, this, &GMainWindow::OnStopGame);
@@ -1087,6 +1083,9 @@ void GMainWindow::ShutdownGame() {
     emulation_running = false;
 
     game_path.clear();
+
+    // When closing the game, destroy the GLWindow to clear the context after the game is closed
+    render_window->ReleaseRenderTarget();
 }
 
 void GMainWindow::StoreRecentFile(const QString& filename) {
@@ -2210,48 +2209,47 @@ void GMainWindow::closeEvent(QCloseEvent* event) {
     QWidget::closeEvent(event);
 }
 
-void GMainWindow::keyPressEvent(QKeyEvent* event) {
-    if (render_window) {
-        render_window->ForwardKeyPressEvent(event);
-    }
+static bool IsSingleFileDropEvent(const QMimeData* mime) {
+    return mime->hasUrls() && mime->urls().length() == 1;
 }
 
-void GMainWindow::keyReleaseEvent(QKeyEvent* event) {
-    if (render_window) {
-        render_window->ForwardKeyReleaseEvent(event);
+void GMainWindow::AcceptDropEvent(QDropEvent* event) {
+    if (IsSingleFileDropEvent(event->mimeData())) {
+        event->setDropAction(Qt::DropAction::LinkAction);
+        event->accept();
     }
 }
 
-static bool IsSingleFileDropEvent(QDropEvent* event) {
-    const QMimeData* mimeData = event->mimeData();
-    return mimeData->hasUrls() && mimeData->urls().length() == 1;
-}
-
-void GMainWindow::dropEvent(QDropEvent* event) {
-    if (!IsSingleFileDropEvent(event)) {
-        return;
+bool GMainWindow::DropAction(QDropEvent* event) {
+    if (!IsSingleFileDropEvent(event->mimeData())) {
+        return false;
     }
 
     const QMimeData* mime_data = event->mimeData();
-    const QString filename = mime_data->urls().at(0).toLocalFile();
+    const QString& filename = mime_data->urls().at(0).toLocalFile();
 
     if (emulation_running && QFileInfo(filename).suffix() == QStringLiteral("bin")) {
+        // Amiibo
         LoadAmiibo(filename);
     } else {
+        // Game
         if (ConfirmChangeGame()) {
             BootGame(filename);
         }
     }
+    return true;
+}
+
+void GMainWindow::dropEvent(QDropEvent* event) {
+    DropAction(event);
 }
 
 void GMainWindow::dragEnterEvent(QDragEnterEvent* event) {
-    if (IsSingleFileDropEvent(event)) {
-        event->acceptProposedAction();
-    }
+    AcceptDropEvent(event);
 }
 
 void GMainWindow::dragMoveEvent(QDragMoveEvent* event) {
-    event->acceptProposedAction();
+    AcceptDropEvent(event);
 }
 
 bool GMainWindow::ConfirmChangeGame() {
@@ -2372,6 +2370,7 @@ int main(int argc, char* argv[]) {
 
     // Enables the core to make the qt created contexts current on std::threads
     QCoreApplication::setAttribute(Qt::AA_DontCheckOpenGLContextThreadAffinity);
+    QCoreApplication::setAttribute(Qt::AA_ShareOpenGLContexts);
     QApplication app(argc, argv);
 
     // Qt changes the locale and causes issues in float conversion using std::to_string() when
diff --git a/src/yuzu/main.h b/src/yuzu/main.h
index 8eba2172c..a67125567 100644
--- a/src/yuzu/main.h
+++ b/src/yuzu/main.h
@@ -78,6 +78,9 @@ public:
 
     std::unique_ptr<DiscordRPC::DiscordInterface> discord_rpc;
 
+    bool DropAction(QDropEvent* event);
+    void AcceptDropEvent(QDropEvent* event);
+
 signals:
 
     /**
@@ -264,8 +267,4 @@ protected:
     void dropEvent(QDropEvent* event) override;
     void dragEnterEvent(QDragEnterEvent* event) override;
     void dragMoveEvent(QDragMoveEvent* event) override;
-
-    // Overrides used to forward signals to the render window when the focus moves out.
-    void keyPressEvent(QKeyEvent* event) override;
-    void keyReleaseEvent(QKeyEvent* event) override;
 };
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index 96f1ce3af..f4cd905c9 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -381,6 +381,8 @@ void Config::ReadValues() {
         static_cast<float>(sdl2_config->GetReal("Renderer", "resolution_factor", 1.0));
     Settings::values.aspect_ratio =
         static_cast<int>(sdl2_config->GetInteger("Renderer", "aspect_ratio", 0));
+    Settings::values.max_anisotropy =
+        static_cast<int>(sdl2_config->GetInteger("Renderer", "max_anisotropy", 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));
@@ -390,6 +392,8 @@ void Config::ReadValues() {
         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.use_vsync =
+        static_cast<u16>(sdl2_config->GetInteger("Renderer", "use_vsync", 1));
 
     Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
     Settings::values.bg_green =
@@ -448,7 +452,7 @@ void Config::ReadValues() {
     Settings::values.yuzu_token = sdl2_config->Get("WebService", "yuzu_token", "");
 
     // Services
-    Settings::values.bcat_backend = sdl2_config->Get("Services", "bcat_backend", "boxcat");
+    Settings::values.bcat_backend = sdl2_config->Get("Services", "bcat_backend", "null");
     Settings::values.bcat_boxcat_local =
         sdl2_config->GetBoolean("Services", "bcat_boxcat_local", false);
 }
diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h
index 8a2b658cd..d63d7a58e 100644
--- a/src/yuzu_cmd/default_ini.h
+++ b/src/yuzu_cmd/default_ini.h
@@ -84,7 +84,7 @@ touch_device=
 # from any cemuhook compatible motion program.
 
 # IPv4 address of the udp input server (Default "127.0.0.1")
-udp_input_address=
+udp_input_address=127.0.0.1
 
 # Port of the udp input server. (Default 26760)
 udp_input_port=
@@ -126,6 +126,10 @@ resolution_factor =
 # 0: Default (16:9), 1: Force 4:3, 2: Force 21:9, 3: Stretch to Window
 aspect_ratio =
 
+# Anisotropic filtering
+# 0: Default, 1: 2x, 2: 4x, 3: 8x, 4: 16x
+max_anisotropy =
+
 # Whether to enable V-Sync (caps the framerate at 60FPS) or not.
 # 0 (default): Off, 1: On
 use_vsync =
@@ -150,6 +154,11 @@ use_accurate_gpu_emulation =
 # 0 : Off (slow), 1 (default): On (fast)
 use_asynchronous_gpu_emulation =
 
+# Forces VSync on the display thread. Usually doesn't impact performance, but on some drivers it can
+# so only turn this off if you notice a speed difference.
+# 0: Off, 1 (default): On
+use_vsync =
+
 # 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 e96139885..19584360c 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
@@ -13,7 +13,7 @@
 #include "input_common/sdl/sdl.h"
 #include "yuzu_cmd/emu_window/emu_window_sdl2.h"
 
-EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
+EmuWindow_SDL2::EmuWindow_SDL2(Core::System& system, bool fullscreen) : system{system} {
     if (SDL_Init(SDL_INIT_VIDEO | SDL_INIT_JOYSTICK) < 0) {
         LOG_CRITICAL(Frontend, "Failed to initialize SDL2! Exiting...");
         exit(1);
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.h b/src/yuzu_cmd/emu_window/emu_window_sdl2.h
index b38f56661..fffac4252 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2.h
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.h
@@ -10,9 +10,13 @@
 
 struct SDL_Window;
 
+namespace Core {
+class System;
+}
+
 class EmuWindow_SDL2 : public Core::Frontend::EmuWindow {
 public:
-    explicit EmuWindow_SDL2(bool fullscreen);
+    explicit EmuWindow_SDL2(Core::System& system, bool fullscreen);
     ~EmuWindow_SDL2();
 
     /// Polls window events
@@ -24,6 +28,9 @@ public:
     /// Returns if window is shown (not minimized)
     bool IsShown() const override;
 
+    /// Presents the next frame
+    virtual void Present() = 0;
+
 protected:
     /// Called by PollEvents when a key is pressed or released.
     void OnKeyEvent(int key, u8 state);
@@ -55,6 +62,9 @@ protected:
     /// Called when a configuration change affects the minimal size of the window
     void OnMinimalClientAreaChangeRequest(std::pair<unsigned, unsigned> minimal_size) override;
 
+    /// Instance of the system, used to access renderer for the presentation thread
+    Core::System& system;
+
     /// Is the window still open?
     bool is_open = true;
 
@@ -62,7 +72,7 @@ protected:
     bool is_shown = true;
 
     /// Internal SDL2 render window
-    SDL_Window* render_window;
+    SDL_Window* render_window{};
 
     /// Keeps track of how often to update the title bar during gameplay
     u32 last_time = 0;
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp
index 7ffa0ac09..c0d373477 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp
@@ -13,24 +13,25 @@
 #include "common/logging/log.h"
 #include "common/scm_rev.h"
 #include "common/string_util.h"
+#include "core/core.h"
 #include "core/settings.h"
 #include "input_common/keyboard.h"
 #include "input_common/main.h"
 #include "input_common/motion_emu.h"
+#include "video_core/renderer_base.h"
 #include "yuzu_cmd/emu_window/emu_window_sdl2_gl.h"
 
 class SDLGLContext : public Core::Frontend::GraphicsContext {
 public:
     explicit SDLGLContext() {
         // create a hidden window to make the shared context against
-        window = SDL_CreateWindow("", SDL_WINDOWPOS_UNDEFINED, // x position
-                                  SDL_WINDOWPOS_UNDEFINED,     // y position
-                                  Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height,
-                                  SDL_WINDOW_OPENGL | SDL_WINDOW_HIDDEN);
+        window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
+                                  SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
         context = SDL_GL_CreateContext(window);
     }
 
     ~SDLGLContext() {
+        DoneCurrent();
         SDL_GL_DeleteContext(context);
         SDL_DestroyWindow(window);
     }
@@ -43,8 +44,6 @@ public:
         SDL_GL_MakeCurrent(window, nullptr);
     }
 
-    void SwapBuffers() override {}
-
 private:
     SDL_Window* window;
     SDL_GLContext context;
@@ -80,7 +79,8 @@ bool EmuWindow_SDL2_GL::SupportsRequiredGLExtensions() {
     return unsupported_ext.empty();
 }
 
-EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscreen) {
+EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(Core::System& system, bool fullscreen)
+    : EmuWindow_SDL2{system, fullscreen} {
     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_COMPATIBILITY);
@@ -90,6 +90,7 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscree
     SDL_GL_SetAttribute(SDL_GL_BLUE_SIZE, 8);
     SDL_GL_SetAttribute(SDL_GL_ALPHA_SIZE, 0);
     SDL_GL_SetAttribute(SDL_GL_SHARE_WITH_CURRENT_CONTEXT, 1);
+    SDL_GL_SetSwapInterval(0);
 
     std::string window_title = fmt::format("yuzu {} | {}-{}", Common::g_build_fullname,
                                            Common::g_scm_branch, Common::g_scm_desc);
@@ -105,13 +106,22 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscree
         exit(1);
     }
 
+    dummy_window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
+                                    SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
+
     if (fullscreen) {
         Fullscreen();
     }
-    gl_context = SDL_GL_CreateContext(render_window);
 
-    if (gl_context == nullptr) {
-        LOG_CRITICAL(Frontend, "Failed to create SDL2 GL context! {}", SDL_GetError());
+    window_context = SDL_GL_CreateContext(render_window);
+    core_context = CreateSharedContext();
+
+    if (window_context == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create SDL2 GL context: {}", SDL_GetError());
+        exit(1);
+    }
+    if (core_context == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create shared SDL2 GL context: {}", SDL_GetError());
         exit(1);
     }
 
@@ -128,28 +138,22 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscree
     OnResize();
     OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
     SDL_PumpEvents();
-    SDL_GL_SetSwapInterval(false);
     LOG_INFO(Frontend, "yuzu Version: {} | {}-{}", Common::g_build_fullname, Common::g_scm_branch,
              Common::g_scm_desc);
     Settings::LogSettings();
-
-    DoneCurrent();
 }
 
 EmuWindow_SDL2_GL::~EmuWindow_SDL2_GL() {
-    SDL_GL_DeleteContext(gl_context);
-}
-
-void EmuWindow_SDL2_GL::SwapBuffers() {
-    SDL_GL_SwapWindow(render_window);
+    core_context.reset();
+    SDL_GL_DeleteContext(window_context);
 }
 
 void EmuWindow_SDL2_GL::MakeCurrent() {
-    SDL_GL_MakeCurrent(render_window, gl_context);
+    core_context->MakeCurrent();
 }
 
 void EmuWindow_SDL2_GL::DoneCurrent() {
-    SDL_GL_MakeCurrent(render_window, nullptr);
+    core_context->DoneCurrent();
 }
 
 void EmuWindow_SDL2_GL::RetrieveVulkanHandlers(void* get_instance_proc_addr, void* instance,
@@ -161,3 +165,13 @@ void EmuWindow_SDL2_GL::RetrieveVulkanHandlers(void* get_instance_proc_addr, voi
 std::unique_ptr<Core::Frontend::GraphicsContext> EmuWindow_SDL2_GL::CreateSharedContext() const {
     return std::make_unique<SDLGLContext>();
 }
+
+void EmuWindow_SDL2_GL::Present() {
+    SDL_GL_MakeCurrent(render_window, window_context);
+    SDL_GL_SetSwapInterval(Settings::values.use_vsync ? 1 : 0);
+    while (IsOpen()) {
+        system.Renderer().TryPresent(100);
+        SDL_GL_SwapWindow(render_window);
+    }
+    SDL_GL_MakeCurrent(render_window, nullptr);
+}
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h
index c753085a8..b80669ff0 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h
@@ -10,17 +10,12 @@
 
 class EmuWindow_SDL2_GL final : public EmuWindow_SDL2 {
 public:
-    explicit EmuWindow_SDL2_GL(bool fullscreen);
+    explicit EmuWindow_SDL2_GL(Core::System& system, bool fullscreen);
     ~EmuWindow_SDL2_GL();
 
-    /// Swap buffers to display the next frame
-    void SwapBuffers() override;
-
-    /// Makes the graphics context current for the caller thread
     void MakeCurrent() override;
-
-    /// Releases the GL context from the caller thread
     void DoneCurrent() override;
+    void Present() override;
 
     /// Ignored in OpenGL
     void RetrieveVulkanHandlers(void* get_instance_proc_addr, void* instance,
@@ -29,10 +24,17 @@ public:
     std::unique_ptr<Core::Frontend::GraphicsContext> CreateSharedContext() const override;
 
 private:
+    /// Fake hidden window for the core context
+    SDL_Window* dummy_window{};
+
     /// Whether the GPU and driver supports the OpenGL extension required
     bool SupportsRequiredGLExtensions();
 
     using SDL_GLContext = void*;
+
     /// The OpenGL context associated with the window
-    SDL_GLContext gl_context;
+    SDL_GLContext window_context;
+
+    /// The OpenGL context associated with the core
+    std::unique_ptr<Core::Frontend::GraphicsContext> core_context;
 };
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
index a203f0da9..abcc58165 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
@@ -15,7 +15,8 @@
 #include "core/settings.h"
 #include "yuzu_cmd/emu_window/emu_window_sdl2_vk.h"
 
-EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(bool fullscreen) : EmuWindow_SDL2(fullscreen) {
+EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(Core::System& system, bool fullscreen)
+    : EmuWindow_SDL2{system, fullscreen} {
     if (SDL_Vulkan_LoadLibrary(nullptr) != 0) {
         LOG_CRITICAL(Frontend, "SDL failed to load the Vulkan library: {}", SDL_GetError());
         exit(EXIT_FAILURE);
@@ -110,8 +111,6 @@ EmuWindow_SDL2_VK::~EmuWindow_SDL2_VK() {
     vkDestroyInstance(vk_instance, nullptr);
 }
 
-void EmuWindow_SDL2_VK::SwapBuffers() {}
-
 void EmuWindow_SDL2_VK::MakeCurrent() {
     // Unused on Vulkan
 }
@@ -160,3 +159,7 @@ bool EmuWindow_SDL2_VK::UseStandardLayers(PFN_vkGetInstanceProcAddr vkGetInstanc
                return layer.layerName == std::string("VK_LAYER_LUNARG_standard_validation");
            }) != layers.end();
 }
+
+void EmuWindow_SDL2_VK::Present() {
+    // TODO (bunnei): ImplementMe
+}
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
index 2a7c06a24..1eb8c0868 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
@@ -10,19 +10,12 @@
 
 class EmuWindow_SDL2_VK final : public EmuWindow_SDL2 {
 public:
-    explicit EmuWindow_SDL2_VK(bool fullscreen);
+    explicit EmuWindow_SDL2_VK(Core::System& system, bool fullscreen);
     ~EmuWindow_SDL2_VK();
 
-    /// Swap buffers to display the next frame
-    void SwapBuffers() override;
-
-    /// Makes the graphics context current for the caller thread
     void MakeCurrent() override;
-
-    /// Releases the GL context from the caller thread
     void DoneCurrent() override;
-
-    /// Retrieves Vulkan specific handlers from the window
+    void Present() override;
     void RetrieveVulkanHandlers(void* get_instance_proc_addr, void* instance,
                                 void* surface) const override;
 
diff --git a/src/yuzu_cmd/yuzu.cpp b/src/yuzu_cmd/yuzu.cpp
index 325795321..babf4c3a4 100644
--- a/src/yuzu_cmd/yuzu.cpp
+++ b/src/yuzu_cmd/yuzu.cpp
@@ -177,14 +177,16 @@ int main(int argc, char** argv) {
     Settings::values.use_gdbstub = use_gdbstub;
     Settings::Apply();
 
+    Core::System& system{Core::System::GetInstance()};
+
     std::unique_ptr<EmuWindow_SDL2> emu_window;
     switch (Settings::values.renderer_backend) {
     case Settings::RendererBackend::OpenGL:
-        emu_window = std::make_unique<EmuWindow_SDL2_GL>(fullscreen);
+        emu_window = std::make_unique<EmuWindow_SDL2_GL>(system, fullscreen);
         break;
     case Settings::RendererBackend::Vulkan:
 #ifdef HAS_VULKAN
-        emu_window = std::make_unique<EmuWindow_SDL2_VK>(fullscreen);
+        emu_window = std::make_unique<EmuWindow_SDL2_VK>(system, fullscreen);
         break;
 #else
         LOG_CRITICAL(Frontend, "Vulkan backend has not been compiled!");
@@ -192,12 +194,6 @@ int main(int argc, char** argv) {
 #endif
     }
 
-    if (!Settings::values.use_multi_core) {
-        // Single core mode must acquire OpenGL context for entire emulation session
-        emu_window->MakeCurrent();
-    }
-
-    Core::System& system{Core::System::GetInstance()};
     system.SetContentProvider(std::make_unique<FileSys::ContentProviderUnion>());
     system.SetFilesystem(std::make_shared<FileSys::RealVfsFilesystem>());
     system.GetFileSystemController().CreateFactories(*system.GetFilesystem());
@@ -234,12 +230,23 @@ int main(int argc, char** argv) {
 
     system.TelemetrySession().AddField(Telemetry::FieldType::App, "Frontend", "SDL");
 
-    emu_window->MakeCurrent();
     system.Renderer().Rasterizer().LoadDiskResources();
 
+    // Acquire render context for duration of the thread if this is the rendering thread
+    if (!Settings::values.use_asynchronous_gpu_emulation) {
+        emu_window->MakeCurrent();
+    }
+    SCOPE_EXIT({
+        if (!Settings::values.use_asynchronous_gpu_emulation) {
+            emu_window->DoneCurrent();
+        }
+    });
+
+    std::thread render_thread([&emu_window] { emu_window->Present(); });
     while (emu_window->IsOpen()) {
         system.RunLoop();
     }
+    render_thread.join();
 
     system.Shutdown();
 
diff --git a/src/yuzu_tester/config.cpp b/src/yuzu_tester/config.cpp
index 0ac93b62a..ee2591c8f 100644
--- a/src/yuzu_tester/config.cpp
+++ b/src/yuzu_tester/config.cpp
@@ -120,6 +120,8 @@ void Config::ReadValues() {
         static_cast<float>(sdl2_config->GetReal("Renderer", "resolution_factor", 1.0));
     Settings::values.aspect_ratio =
         static_cast<int>(sdl2_config->GetInteger("Renderer", "aspect_ratio", 0));
+    Settings::values.max_anisotropy =
+        static_cast<int>(sdl2_config->GetInteger("Renderer", "max_anisotropy", 0));
     Settings::values.use_frame_limit = false;
     Settings::values.frame_limit = 100;
     Settings::values.use_disk_shader_cache =
diff --git a/src/yuzu_tester/default_ini.h b/src/yuzu_tester/default_ini.h
index 8d93f7b88..ca203b64d 100644
--- a/src/yuzu_tester/default_ini.h
+++ b/src/yuzu_tester/default_ini.h
@@ -30,6 +30,10 @@ resolution_factor =
 # 0: Default (16:9), 1: Force 4:3, 2: Force 21:9, 3: Stretch to Window
 aspect_ratio =
 
+# Anisotropic filtering
+# 0: Default, 1: 2x, 2: 4x, 3: 8x, 4: 16x
+max_anisotropy =
+
 # Whether to enable V-Sync (caps the framerate at 60FPS) or not.
 # 0 (default): Off, 1: On
 use_vsync =
diff --git a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp
index f2cc4a797..a1bdb1a12 100644
--- a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp
+++ b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp
@@ -112,10 +112,6 @@ EmuWindow_SDL2_Hide::~EmuWindow_SDL2_Hide() {
     SDL_Quit();
 }
 
-void EmuWindow_SDL2_Hide::SwapBuffers() {
-    SDL_GL_SwapWindow(render_window);
-}
-
 void EmuWindow_SDL2_Hide::PollEvents() {}
 
 void EmuWindow_SDL2_Hide::MakeCurrent() {
diff --git a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h
index c7fccc002..b13e15309 100644
--- a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h
+++ b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h
@@ -13,9 +13,6 @@ public:
     explicit EmuWindow_SDL2_Hide();
     ~EmuWindow_SDL2_Hide();
 
-    /// Swap buffers to display the next frame
-    void SwapBuffers() override;
-
     /// Polls window events
     void PollEvents() override;