50 files changed, 640 insertions, 267 deletions

diff --git a/.travis.yml b/.travis.yml
index 9512f7843..93fda1dfa 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -24,7 +24,7 @@ matrix:
     - os: osx
       env: NAME="macos build"
       sudo: false
-      osx_image: xcode10.1
+      osx_image: xcode10.2
       install: "./.travis/macos/deps.sh"
       script: "./.travis/macos/build.sh"
       after_success: "./.travis/macos/upload.sh"
diff --git a/.travis/linux-mingw/docker.sh b/.travis/linux-mingw/docker.sh
index b73a28693..28033acfb 100755
--- a/.travis/linux-mingw/docker.sh
+++ b/.travis/linux-mingw/docker.sh
@@ -13,8 +13,8 @@ echo '' >> /bin/cmd
 chmod +x /bin/cmd
 
 mkdir build && cd build
-cmake .. -DCMAKE_TOOLCHAIN_FILE="$(pwd)/../CMakeModules/MinGWCross.cmake" -DUSE_CCACHE=ON -DYUZU_USE_BUNDLED_UNICORN=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DCMAKE_BUILD_TYPE=Release
-make -j4
+cmake .. -G Ninja -DCMAKE_TOOLCHAIN_FILE="$(pwd)/../CMakeModules/MinGWCross.cmake" -DUSE_CCACHE=ON -DYUZU_USE_BUNDLED_UNICORN=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DCMAKE_BUILD_TYPE=Release
+ninja
 
 # Clean up the dirty hacks
 rm /bin/uname && mv /bin/uname1 /bin/uname
diff --git a/.travis/linux/docker.sh b/.travis/linux/docker.sh
index a21a8f9e2..3a9970384 100755
--- a/.travis/linux/docker.sh
+++ b/.travis/linux/docker.sh
@@ -3,7 +3,7 @@
 cd /yuzu
 
 mkdir build && cd build
-cmake .. -DYUZU_USE_BUNDLED_UNICORN=ON -DYUZU_USE_QT_WEB_ENGINE=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=/usr/lib/ccache/gcc -DCMAKE_CXX_COMPILER=/usr/lib/ccache/g++ -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DUSE_DISCORD_PRESENCE=ON -G Ninja
+cmake .. -G Ninja -DYUZU_USE_BUNDLED_UNICORN=ON -DYUZU_USE_QT_WEB_ENGINE=ON -DCMAKE_BUILD_TYPE=Release -DCMAKE_C_COMPILER=/usr/lib/ccache/gcc -DCMAKE_CXX_COMPILER=/usr/lib/ccache/g++ -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DUSE_DISCORD_PRESENCE=ON
 ninja
 
 ccache -s
diff --git a/.travis/macos/build.sh b/.travis/macos/build.sh
index b7b4c6f8c..0abd1a93a 100755
--- a/.travis/macos/build.sh
+++ b/.travis/macos/build.sh
@@ -7,6 +7,7 @@ export Qt5_DIR=$(brew --prefix)/opt/qt5
 export UNICORNDIR=$(pwd)/externals/unicorn
 export PATH="/usr/local/opt/ccache/libexec:$PATH"
 
+# TODO: Build using ninja instead of make
 mkdir build && cd build
 cmake --version
 cmake .. -DYUZU_USE_BUNDLED_UNICORN=ON -DYUZU_USE_QT_WEB_ENGINE=ON -DCMAKE_BUILD_TYPE=Release -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${ENABLE_COMPATIBILITY_REPORTING:-"OFF"} -DUSE_DISCORD_PRESENCE=ON
diff --git a/externals/CMakeLists.txt b/externals/CMakeLists.txt
index 3f8b6cda8..e6fa11a03 100644
--- a/externals/CMakeLists.txt
+++ b/externals/CMakeLists.txt
@@ -7,6 +7,10 @@ include(DownloadExternals)
 add_library(catch-single-include INTERFACE)
 target_include_directories(catch-single-include INTERFACE catch/single_include)
 
+# libfmt
+add_subdirectory(fmt)
+add_library(fmt::fmt ALIAS fmt)
+
 # Dynarmic
 if (ARCHITECTURE_x86_64)
     set(DYNARMIC_TESTS OFF)
@@ -14,10 +18,6 @@ if (ARCHITECTURE_x86_64)
     add_subdirectory(dynarmic)
 endif()
 
-# libfmt
-add_subdirectory(fmt)
-add_library(fmt::fmt ALIAS fmt)
-
 # getopt
 if (MSVC)
     add_subdirectory(getopt)
diff --git a/externals/dynarmic b/externals/dynarmic
-Subproject 4e6848d1c9e8dadc70595c15b5589f8b14aad47
+Subproject 2683a9a3e316b5c3f387bbe6787732b9ff44b8d
diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 9aea4af87..04018233f 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -21,15 +21,29 @@ if (MSVC)
     # Ensure that projects build with Unicode support.
     add_definitions(-DUNICODE -D_UNICODE)
 
-    # /W3 - Level 3 warnings
-    # /MP - Multi-threaded compilation
-    # /Zi - Output debugging information
-    # /Zo - enhanced debug info for optimized builds
-    # /permissive- - enables stricter C++ standards conformance checks
-    # /EHsc - C++-only exception handling semantics
-    # /Zc:throwingNew - let codegen assume `operator new` will never return null
-    # /Zc:inline - let codegen omit inline functions in object files
-    add_compile_options(/W3 /MP /Zi /Zo /permissive- /EHsc /std:c++latest /Zc:throwingNew,inline)
+    # /W3                 - Level 3 warnings
+    # /MP                 - Multi-threaded compilation
+    # /Zi                 - Output debugging information
+    # /Zo                 - Enhanced debug info for optimized builds
+    # /permissive-        - Enables stricter C++ standards conformance checks
+    # /EHsc               - C++-only exception handling semantics
+    # /volatile:iso       - Use strict standards-compliant volatile semantics.
+    # /Zc:externConstexpr - Allow extern constexpr variables to have external linkage, like the standard mandates
+    # /Zc:inline          - Let codegen omit inline functions in object files
+    # /Zc:throwingNew     - Let codegen assume `operator new` (without std::nothrow) will never return null
+    add_compile_options(
+        /W3
+        /MP
+        /Zi
+        /Zo
+        /permissive-
+        /EHsc
+        /std:c++latest
+        /volatile:iso
+        /Zc:externConstexpr
+        /Zc:inline
+        /Zc:throwingNew
+    )
 
     # /GS- - No stack buffer overflow checks
     add_compile_options("$<$<CONFIG:Release>:/GS->")
@@ -37,7 +51,10 @@ if (MSVC)
     set(CMAKE_EXE_LINKER_FLAGS_DEBUG   "/DEBUG /MANIFEST:NO" CACHE STRING "" FORCE)
     set(CMAKE_EXE_LINKER_FLAGS_RELEASE "/DEBUG /MANIFEST:NO /INCREMENTAL:NO /OPT:REF,ICF" CACHE STRING "" FORCE)
 else()
-    add_compile_options("-Wno-attributes")
+    add_compile_options(
+        -Wall
+        -Wno-attributes
+    )
 
     if (APPLE AND CMAKE_CXX_COMPILER_ID STREQUAL Clang)
         add_compile_options("-stdlib=libc++")
diff --git a/src/common/zstd_compression.cpp b/src/common/zstd_compression.cpp
index 60a35c67c..978526492 100644
--- a/src/common/zstd_compression.cpp
+++ b/src/common/zstd_compression.cpp
@@ -2,8 +2,6 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#pragma once
-
 #include <algorithm>
 #include <zstd.h>
 
diff --git a/src/core/frontend/emu_window.cpp b/src/core/frontend/emu_window.cpp
index 1320bbe77..eda466a5d 100644
--- a/src/core/frontend/emu_window.cpp
+++ b/src/core/frontend/emu_window.cpp
@@ -10,6 +10,8 @@
 
 namespace Core::Frontend {
 
+GraphicsContext::~GraphicsContext() = default;
+
 class EmuWindow::TouchState : public Input::Factory<Input::TouchDevice>,
                               public std::enable_shared_from_this<TouchState> {
 public:
diff --git a/src/core/frontend/emu_window.h b/src/core/frontend/emu_window.h
index 70a522556..e2c290dc1 100644
--- a/src/core/frontend/emu_window.h
+++ b/src/core/frontend/emu_window.h
@@ -19,6 +19,8 @@ namespace Core::Frontend {
  */
 class GraphicsContext {
 public:
+    virtual ~GraphicsContext();
+
     /// Makes the graphics context current for the caller thread
     virtual void MakeCurrent() = 0;
 
diff --git a/src/core/hle/service/audio/audctl.cpp b/src/core/hle/service/audio/audctl.cpp
index f43e512e9..6a01d4d29 100644
--- a/src/core/hle/service/audio/audctl.cpp
+++ b/src/core/hle/service/audio/audctl.cpp
@@ -50,7 +50,7 @@ void AudCtl::GetTargetVolumeMin(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Audio, "called.");
 
     // This service function is currently hardcoded on the
-    // actual console to this value (as of 6.0.0).
+    // actual console to this value (as of 8.0.0).
     constexpr s32 target_min_volume = 0;
 
     IPC::ResponseBuilder rb{ctx, 3};
@@ -62,7 +62,7 @@ void AudCtl::GetTargetVolumeMax(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Audio, "called.");
 
     // This service function is currently hardcoded on the
-    // actual console to this value (as of 6.0.0).
+    // actual console to this value (as of 8.0.0).
     constexpr s32 target_max_volume = 15;
 
     IPC::ResponseBuilder rb{ctx, 3};
diff --git a/src/core/memory.h b/src/core/memory.h
index b9fa18b1d..04e2c5f1d 100644
--- a/src/core/memory.h
+++ b/src/core/memory.h
@@ -72,15 +72,6 @@ u8* GetPointer(VAddr vaddr);
 
 std::string ReadCString(VAddr vaddr, std::size_t max_length);
 
-enum class FlushMode {
-    /// Write back modified surfaces to RAM
-    Flush,
-    /// Remove region from the cache
-    Invalidate,
-    /// Write back modified surfaces to RAM, and also remove them from the cache
-    FlushAndInvalidate,
-};
-
 /**
  * Mark each page touching the region as cached.
  */
diff --git a/src/core/telemetry_session.cpp b/src/core/telemetry_session.cpp
index e1db06811..4b17bada5 100644
--- a/src/core/telemetry_session.cpp
+++ b/src/core/telemetry_session.cpp
@@ -102,12 +102,6 @@ bool VerifyLogin(const std::string& username, const std::string& token) {
 }
 
 TelemetrySession::TelemetrySession() {
-#ifdef ENABLE_WEB_SERVICE
-    backend = std::make_unique<WebService::TelemetryJson>(
-        Settings::values.web_api_url, Settings::values.yuzu_username, Settings::values.yuzu_token);
-#else
-    backend = std::make_unique<Telemetry::NullVisitor>();
-#endif
     // Log one-time top-level information
     AddField(Telemetry::FieldType::None, "TelemetryId", GetTelemetryId());
 
@@ -175,9 +169,14 @@ TelemetrySession::~TelemetrySession() {
                                 .count()};
     AddField(Telemetry::FieldType::Session, "Shutdown_Time", shutdown_time);
 
+#ifdef ENABLE_WEB_SERVICE
+    auto backend = std::make_unique<WebService::TelemetryJson>(
+        Settings::values.web_api_url, Settings::values.yuzu_username, Settings::values.yuzu_token);
+#else
+    auto backend = std::make_unique<Telemetry::NullVisitor>();
+#endif
+
     // Complete the session, submitting to web service if necessary
-    // This is just a placeholder to wrap up the session once the core completes and this is
-    // destroyed. This will be moved elsewhere once we are actually doing real I/O with the service.
     field_collection.Accept(*backend);
     if (Settings::values.enable_telemetry)
         backend->Complete();
@@ -186,6 +185,8 @@ TelemetrySession::~TelemetrySession() {
 
 bool TelemetrySession::SubmitTestcase() {
 #ifdef ENABLE_WEB_SERVICE
+    auto backend = std::make_unique<WebService::TelemetryJson>(
+        Settings::values.web_api_url, Settings::values.yuzu_username, Settings::values.yuzu_token);
     field_collection.Accept(*backend);
     return backend->SubmitTestcase();
 #else
diff --git a/src/core/telemetry_session.h b/src/core/telemetry_session.h
index 023612b79..cae5a45a0 100644
--- a/src/core/telemetry_session.h
+++ b/src/core/telemetry_session.h
@@ -39,7 +39,6 @@ public:
 
 private:
     Telemetry::FieldCollection field_collection; ///< Tracks all added fields for the session
-    std::unique_ptr<Telemetry::VisitorInterface> backend; ///< Backend interface that logs fields
 };
 
 /**
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 6821f275d..1e010e4da 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -3,6 +3,8 @@ add_library(video_core STATIC
     dma_pusher.h
     debug_utils/debug_utils.cpp
     debug_utils/debug_utils.h
+    engines/engine_upload.cpp
+    engines/engine_upload.h
     engines/fermi_2d.cpp
     engines/fermi_2d.h
     engines/kepler_compute.cpp
diff --git a/src/video_core/engines/engine_upload.cpp b/src/video_core/engines/engine_upload.cpp
new file mode 100644
index 000000000..f8aa4ff55
--- /dev/null
+++ b/src/video_core/engines/engine_upload.cpp
@@ -0,0 +1,48 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "video_core/engines/engine_upload.h"
+#include "video_core/memory_manager.h"
+#include "video_core/textures/decoders.h"
+
+namespace Tegra::Engines::Upload {
+
+State::State(MemoryManager& memory_manager, Registers& regs)
+    : memory_manager(memory_manager), regs(regs) {}
+
+void State::ProcessExec(const bool is_linear) {
+    write_offset = 0;
+    copy_size = regs.line_length_in * regs.line_count;
+    inner_buffer.resize(copy_size);
+    this->is_linear = is_linear;
+}
+
+void State::ProcessData(const u32 data, const bool is_last_call) {
+    const u32 sub_copy_size = std::min(4U, copy_size - write_offset);
+    std::memcpy(&inner_buffer[write_offset], &data, sub_copy_size);
+    write_offset += sub_copy_size;
+    if (!is_last_call) {
+        return;
+    }
+    const GPUVAddr address{regs.dest.Address()};
+    if (is_linear) {
+        memory_manager.WriteBlock(address, inner_buffer.data(), copy_size);
+    } else {
+        UNIMPLEMENTED_IF(regs.dest.z != 0);
+        UNIMPLEMENTED_IF(regs.dest.depth != 1);
+        UNIMPLEMENTED_IF(regs.dest.BlockWidth() != 1);
+        UNIMPLEMENTED_IF(regs.dest.BlockDepth() != 1);
+        const std::size_t dst_size = Tegra::Texture::CalculateSize(
+            true, 1, regs.dest.width, regs.dest.height, 1, regs.dest.BlockHeight(), 1);
+        tmp_buffer.resize(dst_size);
+        memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size);
+        Tegra::Texture::SwizzleKepler(regs.dest.width, regs.dest.height, regs.dest.x, regs.dest.y,
+                                      regs.dest.BlockHeight(), copy_size, inner_buffer.data(),
+                                      tmp_buffer.data());
+        memory_manager.WriteBlock(address, tmp_buffer.data(), dst_size);
+    }
+}
+
+} // namespace Tegra::Engines::Upload
diff --git a/src/video_core/engines/engine_upload.h b/src/video_core/engines/engine_upload.h
new file mode 100644
index 000000000..9c6e0d21c
--- /dev/null
+++ b/src/video_core/engines/engine_upload.h
@@ -0,0 +1,75 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <vector>
+#include "common/bit_field.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Tegra {
+class MemoryManager;
+}
+
+namespace Tegra::Engines::Upload {
+
+struct Registers {
+    u32 line_length_in;
+    u32 line_count;
+
+    struct {
+        u32 address_high;
+        u32 address_low;
+        u32 pitch;
+        union {
+            BitField<0, 4, u32> block_width;
+            BitField<4, 4, u32> block_height;
+            BitField<8, 4, u32> block_depth;
+        };
+        u32 width;
+        u32 height;
+        u32 depth;
+        u32 z;
+        u32 x;
+        u32 y;
+
+        GPUVAddr Address() const {
+            return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) | address_low);
+        }
+
+        u32 BlockWidth() const {
+            return 1U << block_width.Value();
+        }
+
+        u32 BlockHeight() const {
+            return 1U << block_height.Value();
+        }
+
+        u32 BlockDepth() const {
+            return 1U << block_depth.Value();
+        }
+    } dest;
+};
+
+class State {
+public:
+    State(MemoryManager& memory_manager, Registers& regs);
+    ~State() = default;
+
+    void ProcessExec(const bool is_linear);
+    void ProcessData(const u32 data, const bool is_last_call);
+
+private:
+    u32 write_offset = 0;
+    u32 copy_size = 0;
+    std::vector<u8> inner_buffer;
+    std::vector<u8> tmp_buffer;
+    bool is_linear = false;
+    Registers& regs;
+    MemoryManager& memory_manager;
+};
+
+} // namespace Tegra::Engines::Upload
diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h
index 2e51b7f13..45f59a4d9 100644
--- a/src/video_core/engines/fermi_2d.h
+++ b/src/video_core/engines/fermi_2d.h
@@ -21,6 +21,12 @@ class RasterizerInterface;
 
 namespace Tegra::Engines {
 
+/**
+ * This Engine is known as G80_2D. Documentation can be found in:
+ * https://github.com/envytools/envytools/blob/master/rnndb/graph/g80_2d.xml
+ * https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nv50/nv50_2d.xml.h
+ */
+
 #define FERMI2D_REG_INDEX(field_name)                                                              \
     (offsetof(Tegra::Engines::Fermi2D::Regs, field_name) / sizeof(u32))
 
diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp
index b1d950460..7404a8163 100644
--- a/src/video_core/engines/kepler_compute.cpp
+++ b/src/video_core/engines/kepler_compute.cpp
@@ -4,12 +4,21 @@
 
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "core/core.h"
 #include "video_core/engines/kepler_compute.h"
+#include "video_core/engines/maxwell_3d.h"
 #include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_base.h"
+#include "video_core/textures/decoders.h"
 
 namespace Tegra::Engines {
 
-KeplerCompute::KeplerCompute(MemoryManager& memory_manager) : memory_manager{memory_manager} {}
+KeplerCompute::KeplerCompute(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                             MemoryManager& memory_manager)
+    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager}, upload_state{
+                                                                                  memory_manager,
+                                                                                  regs.upload} {}
 
 KeplerCompute::~KeplerCompute() = default;
 
@@ -20,14 +29,34 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
     regs.reg_array[method_call.method] = method_call.argument;
 
     switch (method_call.method) {
+    case KEPLER_COMPUTE_REG_INDEX(exec_upload): {
+        upload_state.ProcessExec(regs.exec_upload.linear != 0);
+        break;
+    }
+    case KEPLER_COMPUTE_REG_INDEX(data_upload): {
+        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_flags.OnMemoryWrite();
+        }
+        break;
+    }
     case KEPLER_COMPUTE_REG_INDEX(launch):
-        // Abort execution since compute shaders can be used to alter game memory (e.g. CUDA
-        // kernels)
-        UNREACHABLE_MSG("Compute shaders are not implemented");
+        ProcessLaunch();
         break;
     default:
         break;
     }
 }
 
+void KeplerCompute::ProcessLaunch() {
+
+    const GPUVAddr launch_desc_loc = regs.launch_desc_loc.Address();
+    memory_manager.ReadBlockUnsafe(launch_desc_loc, &launch_description,
+                                   LaunchParams::NUM_LAUNCH_PARAMETERS * sizeof(u32));
+
+    const GPUVAddr code_loc = regs.code_loc.Address() + launch_description.program_start;
+    LOG_WARNING(HW_GPU, "Compute Kernel Execute at Address 0x{:016x}, STUBBED", code_loc);
+}
+
 } // namespace Tegra::Engines
diff --git a/src/video_core/engines/kepler_compute.h b/src/video_core/engines/kepler_compute.h
index fb6cdf432..5250b8d9b 100644
--- a/src/video_core/engines/kepler_compute.h
+++ b/src/video_core/engines/kepler_compute.h
@@ -6,22 +6,40 @@
 
 #include <array>
 #include <cstddef>
+#include <vector>
+#include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
+#include "video_core/engines/engine_upload.h"
 #include "video_core/gpu.h"
 
+namespace Core {
+class System;
+}
+
 namespace Tegra {
 class MemoryManager;
 }
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Tegra::Engines {
 
+/**
+ * This Engine is known as GK104_Compute. Documentation can be found in:
+ * https://github.com/envytools/envytools/blob/master/rnndb/graph/gk104_compute.xml
+ * https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nvc0/nve4_compute.xml.h
+ */
+
 #define KEPLER_COMPUTE_REG_INDEX(field_name)                                                       \
     (offsetof(Tegra::Engines::KeplerCompute::Regs, field_name) / sizeof(u32))
 
 class KeplerCompute final {
 public:
-    explicit KeplerCompute(MemoryManager& memory_manager);
+    explicit KeplerCompute(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                           MemoryManager& memory_manager);
     ~KeplerCompute();
 
     static constexpr std::size_t NumConstBuffers = 8;
@@ -31,30 +49,181 @@ public:
 
         union {
             struct {
-                INSERT_PADDING_WORDS(0xAF);
+                INSERT_PADDING_WORDS(0x60);
+
+                Upload::Registers upload;
+
+                struct {
+                    union {
+                        BitField<0, 1, u32> linear;
+                    };
+                } exec_upload;
+
+                u32 data_upload;
+
+                INSERT_PADDING_WORDS(0x3F);
+
+                struct {
+                    u32 address;
+                    GPUVAddr Address() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address) << 8));
+                    }
+                } launch_desc_loc;
+
+                INSERT_PADDING_WORDS(0x1);
 
                 u32 launch;
 
-                INSERT_PADDING_WORDS(0xC48);
+                INSERT_PADDING_WORDS(0x4A7);
+
+                struct {
+                    u32 address_high;
+                    u32 address_low;
+                    u32 limit;
+                    GPUVAddr Address() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+                                                     address_low);
+                    }
+                } tsc;
+
+                INSERT_PADDING_WORDS(0x3);
+
+                struct {
+                    u32 address_high;
+                    u32 address_low;
+                    u32 limit;
+                    GPUVAddr Address() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+                                                     address_low);
+                    }
+                } tic;
+
+                INSERT_PADDING_WORDS(0x22);
+
+                struct {
+                    u32 address_high;
+                    u32 address_low;
+                    GPUVAddr Address() const {
+                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
+                                                     address_low);
+                    }
+                } code_loc;
+
+                INSERT_PADDING_WORDS(0x3FE);
+
+                u32 texture_const_buffer_index;
+
+                INSERT_PADDING_WORDS(0x374);
             };
             std::array<u32, NUM_REGS> reg_array;
         };
     } regs{};
+
+    struct LaunchParams {
+        static constexpr std::size_t NUM_LAUNCH_PARAMETERS = 0x40;
+
+        INSERT_PADDING_WORDS(0x8);
+
+        u32 program_start;
+
+        INSERT_PADDING_WORDS(0x2);
+
+        BitField<30, 1, u32> linked_tsc;
+
+        BitField<0, 31, u32> grid_dim_x;
+        union {
+            BitField<0, 16, u32> grid_dim_y;
+            BitField<16, 16, u32> grid_dim_z;
+        };
+
+        INSERT_PADDING_WORDS(0x3);
+
+        BitField<0, 16, u32> shared_alloc;
+
+        BitField<0, 31, u32> block_dim_x;
+        union {
+            BitField<0, 16, u32> block_dim_y;
+            BitField<16, 16, u32> block_dim_z;
+        };
+
+        union {
+            BitField<0, 8, u32> const_buffer_enable_mask;
+            BitField<29, 2, u32> cache_layout;
+        } memory_config;
+
+        INSERT_PADDING_WORDS(0x8);
+
+        struct {
+            u32 address_low;
+            union {
+                BitField<0, 8, u32> address_high;
+                BitField<15, 17, u32> size;
+            };
+            GPUVAddr Address() const {
+                return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high.Value()) << 32) |
+                                             address_low);
+            }
+        } const_buffer_config[8];
+
+        union {
+            BitField<0, 20, u32> local_pos_alloc;
+            BitField<27, 5, u32> barrier_alloc;
+        };
+
+        union {
+            BitField<0, 20, u32> local_neg_alloc;
+            BitField<24, 5, u32> gpr_alloc;
+        };
+
+        INSERT_PADDING_WORDS(0x11);
+    } launch_description;
+
+    struct {
+        u32 write_offset = 0;
+        u32 copy_size = 0;
+        std::vector<u8> inner_buffer;
+    } state{};
+
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32),
                   "KeplerCompute Regs has wrong size");
 
+    static_assert(sizeof(LaunchParams) == LaunchParams::NUM_LAUNCH_PARAMETERS * sizeof(u32),
+                  "KeplerCompute LaunchParams has wrong size");
+
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
 private:
+    Core::System& system;
+    VideoCore::RasterizerInterface& rasterizer;
     MemoryManager& memory_manager;
+    Upload::State upload_state;
+
+    void ProcessLaunch();
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
     static_assert(offsetof(KeplerCompute::Regs, field_name) == position * 4,                       \
                   "Field " #field_name " has invalid position")
 
+#define ASSERT_LAUNCH_PARAM_POSITION(field_name, position)                                         \
+    static_assert(offsetof(KeplerCompute::LaunchParams, field_name) == position * 4,               \
+                  "Field " #field_name " has invalid position")
+
+ASSERT_REG_POSITION(upload, 0x60);
+ASSERT_REG_POSITION(exec_upload, 0x6C);
+ASSERT_REG_POSITION(data_upload, 0x6D);
 ASSERT_REG_POSITION(launch, 0xAF);
+ASSERT_REG_POSITION(tsc, 0x557);
+ASSERT_REG_POSITION(tic, 0x55D);
+ASSERT_REG_POSITION(code_loc, 0x582);
+ASSERT_REG_POSITION(texture_const_buffer_index, 0x982);
+ASSERT_LAUNCH_PARAM_POSITION(program_start, 0x8);
+ASSERT_LAUNCH_PARAM_POSITION(grid_dim_x, 0xC);
+ASSERT_LAUNCH_PARAM_POSITION(shared_alloc, 0x11);
+ASSERT_LAUNCH_PARAM_POSITION(block_dim_x, 0x12);
+ASSERT_LAUNCH_PARAM_POSITION(memory_config, 0x14);
+ASSERT_LAUNCH_PARAM_POSITION(const_buffer_config, 0x1D);
 
 #undef ASSERT_REG_POSITION
 
diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp
index 7387886a3..0561f676c 100644
--- a/src/video_core/engines/kepler_memory.cpp
+++ b/src/video_core/engines/kepler_memory.cpp
@@ -14,9 +14,8 @@
 
 namespace Tegra::Engines {
 
-KeplerMemory::KeplerMemory(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
-                           MemoryManager& memory_manager)
-    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager} {}
+KeplerMemory::KeplerMemory(Core::System& system, MemoryManager& memory_manager)
+    : system{system}, memory_manager{memory_manager}, upload_state{memory_manager, regs.upload} {}
 
 KeplerMemory::~KeplerMemory() = default;
 
@@ -28,46 +27,18 @@ void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) {
 
     switch (method_call.method) {
     case KEPLERMEMORY_REG_INDEX(exec): {
-        ProcessExec();
+        upload_state.ProcessExec(regs.exec.linear != 0);
         break;
     }
     case KEPLERMEMORY_REG_INDEX(data): {
-        ProcessData(method_call.argument, method_call.IsLastCall());
+        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_flags.OnMemoryWrite();
+        }
         break;
     }
     }
 }
 
-void KeplerMemory::ProcessExec() {
-    state.write_offset = 0;
-    state.copy_size = regs.line_length_in * regs.line_count;
-    state.inner_buffer.resize(state.copy_size);
-}
-
-void KeplerMemory::ProcessData(u32 data, bool is_last_call) {
-    const u32 sub_copy_size = std::min(4U, state.copy_size - state.write_offset);
-    std::memcpy(&state.inner_buffer[state.write_offset], &regs.data, sub_copy_size);
-    state.write_offset += sub_copy_size;
-    if (is_last_call) {
-        const GPUVAddr address{regs.dest.Address()};
-        if (regs.exec.linear != 0) {
-            memory_manager.WriteBlock(address, state.inner_buffer.data(), state.copy_size);
-        } else {
-            UNIMPLEMENTED_IF(regs.dest.z != 0);
-            UNIMPLEMENTED_IF(regs.dest.depth != 1);
-            UNIMPLEMENTED_IF(regs.dest.BlockWidth() != 1);
-            UNIMPLEMENTED_IF(regs.dest.BlockDepth() != 1);
-            const std::size_t dst_size = Tegra::Texture::CalculateSize(
-                true, 1, regs.dest.width, regs.dest.height, 1, regs.dest.BlockHeight(), 1);
-            std::vector<u8> tmp_buffer(dst_size);
-            memory_manager.ReadBlock(address, tmp_buffer.data(), dst_size);
-            Tegra::Texture::SwizzleKepler(regs.dest.width, regs.dest.height, regs.dest.x,
-                                          regs.dest.y, regs.dest.BlockHeight(), state.copy_size,
-                                          state.inner_buffer.data(), tmp_buffer.data());
-            memory_manager.WriteBlock(address, tmp_buffer.data(), dst_size);
-        }
-        system.GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
-    }
-}
-
 } // namespace Tegra::Engines
diff --git a/src/video_core/engines/kepler_memory.h b/src/video_core/engines/kepler_memory.h
index 5f892ddad..f3bc675a9 100644
--- a/src/video_core/engines/kepler_memory.h
+++ b/src/video_core/engines/kepler_memory.h
@@ -10,6 +10,7 @@
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
+#include "video_core/engines/engine_upload.h"
 #include "video_core/gpu.h"
 
 namespace Core {
@@ -20,19 +21,20 @@ namespace Tegra {
 class MemoryManager;
 }
 
-namespace VideoCore {
-class RasterizerInterface;
-}
-
 namespace Tegra::Engines {
 
+/**
+ * This Engine is known as P2MF. Documentation can be found in:
+ * https://github.com/envytools/envytools/blob/master/rnndb/graph/gk104_p2mf.xml
+ * https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nvc0/nve4_p2mf.xml.h
+ */
+
 #define KEPLERMEMORY_REG_INDEX(field_name)                                                         \
     (offsetof(Tegra::Engines::KeplerMemory::Regs, field_name) / sizeof(u32))
 
 class KeplerMemory final {
 public:
-    KeplerMemory(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
-                 MemoryManager& memory_manager);
+    KeplerMemory(Core::System& system, MemoryManager& memory_manager);
     ~KeplerMemory();
 
     /// Write the value to the register identified by method.
@@ -45,42 +47,7 @@ public:
             struct {
                 INSERT_PADDING_WORDS(0x60);
 
-                u32 line_length_in;
-                u32 line_count;
-
-                struct {
-                    u32 address_high;
-                    u32 address_low;
-                    u32 pitch;
-                    union {
-                        BitField<0, 4, u32> block_width;
-                        BitField<4, 4, u32> block_height;
-                        BitField<8, 4, u32> block_depth;
-                    };
-                    u32 width;
-                    u32 height;
-                    u32 depth;
-                    u32 z;
-                    u32 x;
-                    u32 y;
-
-                    GPUVAddr Address() const {
-                        return static_cast<GPUVAddr>((static_cast<GPUVAddr>(address_high) << 32) |
-                                                     address_low);
-                    }
-
-                    u32 BlockWidth() const {
-                        return 1U << block_width.Value();
-                    }
-
-                    u32 BlockHeight() const {
-                        return 1U << block_height.Value();
-                    }
-
-                    u32 BlockDepth() const {
-                        return 1U << block_depth.Value();
-                    }
-                } dest;
+                Upload::Registers upload;
 
                 struct {
                     union {
@@ -96,28 +63,17 @@ public:
         };
     } regs{};
 
-    struct {
-        u32 write_offset = 0;
-        u32 copy_size = 0;
-        std::vector<u8> inner_buffer;
-    } state{};
-
 private:
     Core::System& system;
-    VideoCore::RasterizerInterface& rasterizer;
     MemoryManager& memory_manager;
-
-    void ProcessExec();
-    void ProcessData(u32 data, bool is_last_call);
+    Upload::State upload_state;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
     static_assert(offsetof(KeplerMemory::Regs, field_name) == position * 4,                        \
                   "Field " #field_name " has invalid position")
 
-ASSERT_REG_POSITION(line_length_in, 0x60);
-ASSERT_REG_POSITION(line_count, 0x61);
-ASSERT_REG_POSITION(dest, 0x62);
+ASSERT_REG_POSITION(upload, 0x60);
 ASSERT_REG_POSITION(exec, 0x6C);
 ASSERT_REG_POSITION(data, 0x6D);
 #undef ASSERT_REG_POSITION
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index 9780417f2..d7b586db9 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -20,8 +20,8 @@ constexpr u32 MacroRegistersStart = 0xE00;
 
 Maxwell3D::Maxwell3D(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                      MemoryManager& memory_manager)
-    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager}, macro_interpreter{
-                                                                                  *this} {
+    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager},
+      macro_interpreter{*this}, upload_state{memory_manager, regs.upload} {
     InitializeRegisterDefaults();
 }
 
@@ -253,6 +253,18 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
         ProcessSyncPoint();
         break;
     }
+    case MAXWELL3D_REG_INDEX(exec_upload): {
+        upload_state.ProcessExec(regs.exec_upload.linear != 0);
+        break;
+    }
+    case MAXWELL3D_REG_INDEX(data_upload): {
+        const bool is_last_call = method_call.IsLastCall();
+        upload_state.ProcessData(method_call.argument, is_last_call);
+        if (is_last_call) {
+            dirty_flags.OnMemoryWrite();
+        }
+        break;
+    }
     default:
         break;
     }
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 85d309d9b..4883b582a 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -14,6 +14,7 @@
 #include "common/common_funcs.h"
 #include "common/common_types.h"
 #include "common/math_util.h"
+#include "video_core/engines/engine_upload.h"
 #include "video_core/gpu.h"
 #include "video_core/macro_interpreter.h"
 #include "video_core/textures/texture.h"
@@ -32,6 +33,12 @@ class RasterizerInterface;
 
 namespace Tegra::Engines {
 
+/**
+ * This Engine is known as GF100_3D. Documentation can be found in:
+ * https://github.com/envytools/envytools/blob/master/rnndb/graph/gf100_3d.xml
+ * https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/drivers/nouveau/nvc0/nvc0_3d.xml.h
+ */
+
 #define MAXWELL3D_REG_INDEX(field_name)                                                            \
     (offsetof(Tegra::Engines::Maxwell3D::Regs, field_name) / sizeof(u32))
 
@@ -580,7 +587,18 @@ public:
                     u32 bind;
                 } macros;
 
-                INSERT_PADDING_WORDS(0x69);
+                INSERT_PADDING_WORDS(0x17);
+
+                Upload::Registers upload;
+                struct {
+                    union {
+                        BitField<0, 1, u32> linear;
+                    };
+                } exec_upload;
+
+                u32 data_upload;
+
+                INSERT_PADDING_WORDS(0x44);
 
                 struct {
                     union {
@@ -1176,6 +1194,8 @@ private:
     /// Interpreter for the macro codes uploaded to the GPU.
     MacroInterpreter macro_interpreter;
 
+    Upload::State upload_state;
+
     /// Retrieves information about a specific TIC entry from the TIC buffer.
     Texture::TICEntry GetTICEntry(u32 tic_index) const;
 
@@ -1219,6 +1239,9 @@ private:
                   "Field " #field_name " has invalid position")
 
 ASSERT_REG_POSITION(macros, 0x45);
+ASSERT_REG_POSITION(upload, 0x60);
+ASSERT_REG_POSITION(exec_upload, 0x6C);
+ASSERT_REG_POSITION(data_upload, 0x6D);
 ASSERT_REG_POSITION(sync_info, 0xB2);
 ASSERT_REG_POSITION(tfb_enabled, 0x1D1);
 ASSERT_REG_POSITION(rt, 0x200);
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp
index 2426d0067..3a5dfef0c 100644
--- a/src/video_core/engines/maxwell_dma.cpp
+++ b/src/video_core/engines/maxwell_dma.cpp
@@ -83,57 +83,66 @@ void MaxwellDMA::HandleCopy() {
 
     ASSERT(regs.exec.enable_2d == 1);
 
-    const std::size_t copy_size = regs.x_count * regs.y_count;
+    if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) {
+        ASSERT(regs.src_params.size_z == 1);
+        // If the input is tiled and the output is linear, deswizzle the input and copy it over.
+        const u32 src_bytes_per_pixel = regs.src_pitch / regs.src_params.size_x;
+        const std::size_t src_size = Texture::CalculateSize(
+            true, src_bytes_per_pixel, regs.src_params.size_x, regs.src_params.size_y,
+            regs.src_params.size_z, regs.src_params.BlockHeight(), regs.src_params.BlockDepth());
 
-    auto source_ptr{memory_manager.GetPointer(source)};
-    auto dst_ptr{memory_manager.GetPointer(dest)};
+        const std::size_t dst_size = regs.dst_pitch * regs.y_count;
 
-    if (!source_ptr) {
-        LOG_ERROR(HW_GPU, "source_ptr is invalid");
-        return;
-    }
+        if (read_buffer.size() < src_size) {
+            read_buffer.resize(src_size);
+        }
 
-    if (!dst_ptr) {
-        LOG_ERROR(HW_GPU, "dst_ptr is invalid");
-        return;
-    }
+        if (write_buffer.size() < dst_size) {
+            write_buffer.resize(dst_size);
+        }
 
-    const auto FlushAndInvalidate = [&](u32 src_size, u64 dst_size) {
-        // TODO(Subv): For now, manually flush the regions until we implement GPU-accelerated
-        // copying.
-        rasterizer.FlushRegion(ToCacheAddr(source_ptr), src_size);
+        memory_manager.ReadBlock(source, read_buffer.data(), src_size);
+        memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
 
-        // We have to invalidate the destination region to evict any outdated surfaces from the
-        // cache. We do this before actually writing the new data because the destination address
-        // might contain a dirty surface that will have to be written back to memory.
-        rasterizer.InvalidateRegion(ToCacheAddr(dst_ptr), dst_size);
-    };
+        Texture::UnswizzleSubrect(regs.x_count, regs.y_count, regs.dst_pitch,
+                                  regs.src_params.size_x, src_bytes_per_pixel, read_buffer.data(),
+                                  write_buffer.data(), regs.src_params.BlockHeight(),
+                                  regs.src_params.pos_x, regs.src_params.pos_y);
 
-    if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) {
-        ASSERT(regs.src_params.size_z == 1);
-        // If the input is tiled and the output is linear, deswizzle the input and copy it over.
+        memory_manager.WriteBlock(dest, write_buffer.data(), dst_size);
+    } else {
+        ASSERT(regs.dst_params.BlockDepth() == 1);
 
-        const u32 src_bytes_per_pixel = regs.src_pitch / regs.src_params.size_x;
+        const u32 src_bytes_per_pixel = regs.src_pitch / regs.x_count;
 
-        FlushAndInvalidate(regs.src_pitch * regs.src_params.size_y,
-                           copy_size * src_bytes_per_pixel);
+        const std::size_t dst_size = Texture::CalculateSize(
+            true, src_bytes_per_pixel, regs.dst_params.size_x, regs.dst_params.size_y,
+            regs.dst_params.size_z, regs.dst_params.BlockHeight(), regs.dst_params.BlockDepth());
 
-        Texture::UnswizzleSubrect(regs.x_count, regs.y_count, regs.dst_pitch,
-                                  regs.src_params.size_x, src_bytes_per_pixel, source_ptr, dst_ptr,
-                                  regs.src_params.BlockHeight(), regs.src_params.pos_x,
-                                  regs.src_params.pos_y);
-    } else {
-        ASSERT(regs.dst_params.size_z == 1);
-        ASSERT(regs.src_pitch == regs.x_count);
+        const std::size_t dst_layer_size = Texture::CalculateSize(
+            true, src_bytes_per_pixel, regs.dst_params.size_x, regs.dst_params.size_y, 1,
+            regs.dst_params.BlockHeight(), regs.dst_params.BlockDepth());
 
-        const u32 src_bpp = regs.src_pitch / regs.x_count;
+        const std::size_t src_size = regs.src_pitch * regs.y_count;
 
-        FlushAndInvalidate(regs.src_pitch * regs.y_count,
-                           regs.dst_params.size_x * regs.dst_params.size_y * src_bpp);
+        if (read_buffer.size() < src_size) {
+            read_buffer.resize(src_size);
+        }
+
+        if (write_buffer.size() < dst_size) {
+            write_buffer.resize(dst_size);
+        }
+
+        memory_manager.ReadBlock(source, read_buffer.data(), src_size);
+        memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
 
         // If the input is linear and the output is tiled, swizzle the input and copy it over.
         Texture::SwizzleSubrect(regs.x_count, regs.y_count, regs.src_pitch, regs.dst_params.size_x,
-                                src_bpp, dst_ptr, source_ptr, regs.dst_params.BlockHeight());
+                                src_bytes_per_pixel,
+                                write_buffer.data() + dst_layer_size * regs.dst_params.pos_z,
+                                read_buffer.data(), regs.dst_params.BlockHeight());
+
+        memory_manager.WriteBlock(dest, write_buffer.data(), dst_size);
     }
 }
 
diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h
index c6b649842..e5942f671 100644
--- a/src/video_core/engines/maxwell_dma.h
+++ b/src/video_core/engines/maxwell_dma.h
@@ -6,6 +6,7 @@
 
 #include <array>
 #include <cstddef>
+#include <vector>
 #include "common/bit_field.h"
 #include "common/common_funcs.h"
 #include "common/common_types.h"
@@ -25,6 +26,11 @@ class RasterizerInterface;
 
 namespace Tegra::Engines {
 
+/**
+ * This Engine is known as GK104_Copy. Documentation can be found in:
+ * https://github.com/envytools/envytools/blob/master/rnndb/fifo/gk104_copy.xml
+ */
+
 class MaxwellDMA final {
 public:
     explicit MaxwellDMA(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
@@ -63,6 +69,16 @@ public:
 
         static_assert(sizeof(Parameters) == 24, "Parameters has wrong size");
 
+        enum class ComponentMode : u32 {
+            Src0 = 0,
+            Src1 = 1,
+            Src2 = 2,
+            Src3 = 3,
+            Const0 = 4,
+            Const1 = 5,
+            Zero = 6,
+        };
+
         enum class CopyMode : u32 {
             None = 0,
             Unk1 = 1,
@@ -128,7 +144,26 @@ public:
                 u32 x_count;
                 u32 y_count;
 
-                INSERT_PADDING_WORDS(0xBB);
+                INSERT_PADDING_WORDS(0xB8);
+
+                u32 const0;
+                u32 const1;
+                union {
+                    BitField<0, 4, ComponentMode> component0;
+                    BitField<4, 4, ComponentMode> component1;
+                    BitField<8, 4, ComponentMode> component2;
+                    BitField<12, 4, ComponentMode> component3;
+                    BitField<16, 2, u32> component_size;
+                    BitField<20, 3, u32> src_num_components;
+                    BitField<24, 3, u32> dst_num_components;
+
+                    u32 SrcBytePerPixel() const {
+                        return src_num_components.Value() * component_size.Value();
+                    }
+                    u32 DstBytePerPixel() const {
+                        return dst_num_components.Value() * component_size.Value();
+                    }
+                } swizzle_config;
 
                 Parameters dst_params;
 
@@ -149,6 +184,9 @@ private:
 
     MemoryManager& memory_manager;
 
+    std::vector<u8> read_buffer;
+    std::vector<u8> write_buffer;
+
     /// Performs the copy from the source buffer to the destination buffer as configured in the
     /// registers.
     void HandleCopy();
@@ -165,6 +203,9 @@ ASSERT_REG_POSITION(src_pitch, 0x104);
 ASSERT_REG_POSITION(dst_pitch, 0x105);
 ASSERT_REG_POSITION(x_count, 0x106);
 ASSERT_REG_POSITION(y_count, 0x107);
+ASSERT_REG_POSITION(const0, 0x1C0);
+ASSERT_REG_POSITION(const1, 0x1C1);
+ASSERT_REG_POSITION(swizzle_config, 0x1C2);
 ASSERT_REG_POSITION(dst_params, 0x1C3);
 ASSERT_REG_POSITION(src_params, 0x1CA);
 
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 4461083ff..52706505b 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -35,9 +35,9 @@ GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer) : renderer{ren
     dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);
     fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer, *memory_manager);
-    kepler_compute = std::make_unique<Engines::KeplerCompute>(*memory_manager);
+    kepler_compute = std::make_unique<Engines::KeplerCompute>(system, rasterizer, *memory_manager);
     maxwell_dma = std::make_unique<Engines::MaxwellDMA>(system, rasterizer, *memory_manager);
-    kepler_memory = std::make_unique<Engines::KeplerMemory>(system, rasterizer, *memory_manager);
+    kepler_memory = std::make_unique<Engines::KeplerMemory>(system, *memory_manager);
 }
 
 GPU::~GPU() = default;
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index c9a2077de..03856013f 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -44,7 +44,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
                 renderer.Rasterizer().FlushRegion(data->addr, data->size);
             } else if (const auto data = std::get_if<InvalidateRegionCommand>(&next.data)) {
                 renderer.Rasterizer().InvalidateRegion(data->addr, data->size);
-            } else if (const auto data = std::get_if<EndProcessingCommand>(&next.data)) {
+            } else if (std::holds_alternative<EndProcessingCommand>(next.data)) {
                 return;
             } else {
                 UNREACHABLE();
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index 6c98c6701..5d8d126c1 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -25,6 +25,8 @@ MemoryManager::MemoryManager(VideoCore::RasterizerInterface& rasterizer) : raste
     UpdatePageTableForVMA(initial_vma);
 }
 
+MemoryManager::~MemoryManager() = default;
+
 GPUVAddr MemoryManager::AllocateSpace(u64 size, u64 align) {
     const u64 aligned_size{Common::AlignUp(size, page_size)};
     const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)};
@@ -199,11 +201,11 @@ const u8* MemoryManager::GetPointer(GPUVAddr addr) const {
     return {};
 }
 
-bool MemoryManager::IsBlockContinous(const GPUVAddr start, const std::size_t size) {
+bool MemoryManager::IsBlockContinuous(const GPUVAddr start, const std::size_t size) const {
     const GPUVAddr end = start + size;
     const auto host_ptr_start = reinterpret_cast<std::uintptr_t>(GetPointer(start));
     const auto host_ptr_end = reinterpret_cast<std::uintptr_t>(GetPointer(end));
-    const std::size_t range = static_cast<std::size_t>(host_ptr_end - host_ptr_start);
+    const auto range = static_cast<std::size_t>(host_ptr_end - host_ptr_start);
     return range == size;
 }
 
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index e4f0c4bd6..113f9d8f3 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -47,7 +47,8 @@ struct VirtualMemoryArea {
 
 class MemoryManager final {
 public:
-    MemoryManager(VideoCore::RasterizerInterface& rasterizer);
+    explicit MemoryManager(VideoCore::RasterizerInterface& rasterizer);
+    ~MemoryManager();
 
     GPUVAddr AllocateSpace(u64 size, u64 align);
     GPUVAddr AllocateSpace(GPUVAddr addr, u64 size, u64 align);
@@ -65,18 +66,18 @@ public:
     u8* GetPointer(GPUVAddr addr);
     const u8* GetPointer(GPUVAddr addr) const;
 
-    // Returns true if the block is continous in host memory, false otherwise
-    bool IsBlockContinous(const GPUVAddr start, const std::size_t size);
+    /// Returns true if the block is continuous in host memory, false otherwise
+    bool IsBlockContinuous(GPUVAddr start, std::size_t size) const;
 
     /**
      * ReadBlock and WriteBlock are full read and write operations over virtual
-     * GPU Memory. It's important to use these when GPU memory may not be continous
+     * GPU Memory. It's important to use these when GPU memory may not be continuous
      * in the Host Memory counterpart. Note: This functions cause Host GPU Memory
      * Flushes and Invalidations, respectively to each operation.
      */
-    void ReadBlock(GPUVAddr src_addr, void* dest_buffer, const std::size_t size) const;
-    void WriteBlock(GPUVAddr dest_addr, const void* src_buffer, const std::size_t size);
-    void CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size);
+    void ReadBlock(GPUVAddr src_addr, void* dest_buffer, std::size_t size) const;
+    void WriteBlock(GPUVAddr dest_addr, const void* src_buffer, std::size_t size);
+    void CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size);
 
     /**
      * ReadBlockUnsafe and WriteBlockUnsafe are special versions of ReadBlock and
@@ -88,9 +89,9 @@ public:
      * WriteBlockUnsafe instead of WriteBlock since it shouldn't invalidate the texture
      * being flushed.
      */
-    void ReadBlockUnsafe(GPUVAddr src_addr, void* dest_buffer, const std::size_t size) const;
-    void WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer, const std::size_t size);
-    void CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, const std::size_t size);
+    void ReadBlockUnsafe(GPUVAddr src_addr, void* dest_buffer, std::size_t size) const;
+    void WriteBlockUnsafe(GPUVAddr dest_addr, const void* src_buffer, std::size_t size);
+    void CopyBlockUnsafe(GPUVAddr dest_addr, GPUVAddr src_addr, std::size_t size);
 
 private:
     using VMAMap = std::map<GPUVAddr, VirtualMemoryArea>;
@@ -111,10 +112,10 @@ private:
     /**
      * Maps an unmanaged host memory pointer at a given address.
      *
-     * @param target The guest address to start the mapping at.
-     * @param memory The memory to be mapped.
-     * @param size Size of the mapping.
-     * @param state MemoryState tag to attach to the VMA.
+     * @param target       The guest address to start the mapping at.
+     * @param memory       The memory to be mapped.
+     * @param size         Size of the mapping in bytes.
+     * @param backing_addr The base address of the range to back this mapping.
      */
     VMAHandle MapBackingMemory(GPUVAddr target, u8* memory, u64 size, VAddr backing_addr);
 
@@ -124,7 +125,7 @@ private:
     /// Converts a VMAHandle to a mutable VMAIter.
     VMAIter StripIterConstness(const VMAHandle& iter);
 
-    /// Marks as the specfied VMA as allocated.
+    /// Marks as the specified VMA as allocated.
     VMAIter Allocate(VMAIter vma);
 
     /**
diff --git a/src/video_core/rasterizer_cache.h b/src/video_core/rasterizer_cache.h
index 291772186..f820f3ed9 100644
--- a/src/video_core/rasterizer_cache.h
+++ b/src/video_core/rasterizer_cache.h
@@ -37,9 +37,6 @@ public:
     /// Gets the size of the shader in guest memory, required for cache management
     virtual std::size_t GetSizeInBytes() const = 0;
 
-    /// Wriets any cached resources back to memory
-    virtual void Flush() = 0;
-
     /// Sets whether the cached object should be considered registered
     void SetIsRegistered(bool registered) {
         is_registered = registered;
@@ -158,6 +155,8 @@ protected:
         return ++modified_ticks;
     }
 
+    virtual void FlushObjectInner(const T& object) = 0;
+
     /// Flushes the specified object, updating appropriate cache state as needed
     void FlushObject(const T& object) {
         std::lock_guard lock{mutex};
@@ -165,7 +164,7 @@ protected:
         if (!object->IsDirty()) {
             return;
         }
-        object->Flush();
+        FlushObjectInner(object);
         object->MarkAsModified(false, *this);
     }
 
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.h b/src/video_core/renderer_opengl/gl_buffer_cache.h
index fc33aa433..f9247a40e 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.h
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.h
@@ -42,9 +42,6 @@ public:
         return alignment;
     }
 
-    // We do not have to flush this cache as things in it are never modified by us.
-    void Flush() override {}
-
 private:
     VAddr cpu_addr{};
     std::size_t size{};
@@ -75,6 +72,9 @@ public:
 protected:
     void AlignBuffer(std::size_t alignment);
 
+    // We do not have to flush this cache as things in it are never modified by us.
+    void FlushObjectInner(const std::shared_ptr<CachedBufferEntry>& object) override {}
+
 private:
     OGLStreamBuffer stream_buffer;
 
diff --git a/src/video_core/renderer_opengl/gl_global_cache.h b/src/video_core/renderer_opengl/gl_global_cache.h
index 196e6e278..2d467a240 100644
--- a/src/video_core/renderer_opengl/gl_global_cache.h
+++ b/src/video_core/renderer_opengl/gl_global_cache.h
@@ -46,7 +46,7 @@ public:
     /// Reloads the global region from guest memory
     void Reload(u32 size_);
 
-    void Flush() override;
+    void Flush();
 
 private:
     VAddr cpu_addr{};
@@ -65,6 +65,11 @@ public:
     GlobalRegion GetGlobalRegion(const GLShader::GlobalMemoryEntry& descriptor,
                                  Tegra::Engines::Maxwell3D::Regs::ShaderStage stage);
 
+protected:
+    void FlushObjectInner(const GlobalRegion& object) override {
+        object->Flush();
+    }
+
 private:
     GlobalRegion TryGetReservedGlobalRegion(CacheAddr addr, u32 size) const;
     GlobalRegion GetUncachedGlobalRegion(GPUVAddr addr, u8* host_ptr, u32 size);
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index db73e746c..3cc945235 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -922,8 +922,8 @@ void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
         viewport.y = viewport_rect.bottom;
         viewport.width = viewport_rect.GetWidth();
         viewport.height = viewport_rect.GetHeight();
-        viewport.depth_range_far = regs.viewports[i].depth_range_far;
-        viewport.depth_range_near = regs.viewports[i].depth_range_near;
+        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;
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 5a25f5b37..a7681902e 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -628,9 +628,11 @@ CachedSurface::CachedSurface(const SurfaceParams& params)
 }
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceLoad, "OpenGL", "Surface Load", MP_RGB(128, 192, 64));
-void CachedSurface::LoadGLBuffer() {
+void CachedSurface::LoadGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem) {
     MICROPROFILE_SCOPE(OpenGL_SurfaceLoad);
-    gl_buffer.resize(params.max_mip_level);
+    auto& gl_buffer = res_cache_tmp_mem.gl_buffer;
+    if (gl_buffer.size() < params.max_mip_level)
+        gl_buffer.resize(params.max_mip_level);
     for (u32 i = 0; i < params.max_mip_level; i++)
         gl_buffer[i].resize(params.GetMipmapSizeGL(i));
     if (params.is_tiled) {
@@ -671,13 +673,13 @@ void CachedSurface::LoadGLBuffer() {
 }
 
 MICROPROFILE_DEFINE(OpenGL_SurfaceFlush, "OpenGL", "Surface Flush", MP_RGB(128, 192, 64));
-void CachedSurface::FlushGLBuffer() {
+void CachedSurface::FlushGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem) {
     MICROPROFILE_SCOPE(OpenGL_SurfaceFlush);
 
     ASSERT_MSG(!IsPixelFormatASTC(params.pixel_format), "Unimplemented");
 
+    auto& gl_buffer = res_cache_tmp_mem.gl_buffer;
     // OpenGL temporary buffer needs to be big enough to store raw texture size
-    gl_buffer.resize(1);
     gl_buffer[0].resize(GetSizeInBytes());
 
     const FormatTuple& tuple = GetFormatTuple(params.pixel_format, params.component_type);
@@ -713,10 +715,12 @@ void CachedSurface::FlushGLBuffer() {
     }
 }
 
-void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
-                                          GLuint draw_fb_handle) {
+void CachedSurface::UploadGLMipmapTexture(RasterizerTemporaryMemory& res_cache_tmp_mem, u32 mip_map,
+                                          GLuint read_fb_handle, GLuint draw_fb_handle) {
     const auto& rect{params.GetRect(mip_map)};
 
+    auto& gl_buffer = res_cache_tmp_mem.gl_buffer;
+
     // Load data from memory to the surface
     const auto x0 = static_cast<GLint>(rect.left);
     const auto y0 = static_cast<GLint>(rect.bottom);
@@ -801,7 +805,6 @@ void CachedSurface::UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle,
                                 tuple.type, &gl_buffer[mip_map][buffer_offset]);
             break;
         case SurfaceTarget::TextureCubemap: {
-            std::size_t start = buffer_offset;
             for (std::size_t face = 0; face < params.depth; ++face) {
                 glTextureSubImage3D(texture.handle, mip_map, x0, y0, static_cast<GLint>(face),
                                     static_cast<GLsizei>(rect.GetWidth()),
@@ -845,11 +848,12 @@ void CachedSurface::EnsureTextureDiscrepantView() {
 }
 
 MICROPROFILE_DEFINE(OpenGL_TextureUL, "OpenGL", "Texture Upload", MP_RGB(128, 192, 64));
-void CachedSurface::UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle) {
+void CachedSurface::UploadGLTexture(RasterizerTemporaryMemory& res_cache_tmp_mem,
+                                    GLuint read_fb_handle, GLuint draw_fb_handle) {
     MICROPROFILE_SCOPE(OpenGL_TextureUL);
 
     for (u32 i = 0; i < params.max_mip_level; i++)
-        UploadGLMipmapTexture(i, read_fb_handle, draw_fb_handle);
+        UploadGLMipmapTexture(res_cache_tmp_mem, i, read_fb_handle, draw_fb_handle);
 }
 
 void CachedSurface::UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
@@ -929,8 +933,8 @@ Surface RasterizerCacheOpenGL::GetColorBufferSurface(std::size_t index, bool pre
 }
 
 void RasterizerCacheOpenGL::LoadSurface(const Surface& surface) {
-    surface->LoadGLBuffer();
-    surface->UploadGLTexture(read_framebuffer.handle, draw_framebuffer.handle);
+    surface->LoadGLBuffer(temporal_memory);
+    surface->UploadGLTexture(temporal_memory, read_framebuffer.handle, draw_framebuffer.handle);
     surface->MarkAsModified(false, *this);
     surface->MarkForReload(false);
 }
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
index db280dbb3..6263ef3e7 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h
@@ -355,6 +355,12 @@ namespace OpenGL {
 
 class RasterizerOpenGL;
 
+// This is used to store temporary big buffers,
+// instead of creating/destroying all the time
+struct RasterizerTemporaryMemory {
+    std::vector<std::vector<u8>> gl_buffer;
+};
+
 class CachedSurface final : public RasterizerCacheObject {
 public:
     explicit CachedSurface(const SurfaceParams& params);
@@ -371,10 +377,6 @@ public:
         return memory_size;
     }
 
-    void Flush() override {
-        FlushGLBuffer();
-    }
-
     const OGLTexture& Texture() const {
         return texture;
     }
@@ -397,11 +399,12 @@ public:
     }
 
     // Read/Write data in Switch memory to/from gl_buffer
-    void LoadGLBuffer();
-    void FlushGLBuffer();
+    void LoadGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem);
+    void FlushGLBuffer(RasterizerTemporaryMemory& res_cache_tmp_mem);
 
     // Upload data in gl_buffer to this surface's texture
-    void UploadGLTexture(GLuint read_fb_handle, GLuint draw_fb_handle);
+    void UploadGLTexture(RasterizerTemporaryMemory& res_cache_tmp_mem, GLuint read_fb_handle,
+                         GLuint draw_fb_handle);
 
     void UpdateSwizzle(Tegra::Texture::SwizzleSource swizzle_x,
                        Tegra::Texture::SwizzleSource swizzle_y,
@@ -429,13 +432,13 @@ public:
     }
 
 private:
-    void UploadGLMipmapTexture(u32 mip_map, GLuint read_fb_handle, GLuint draw_fb_handle);
+    void UploadGLMipmapTexture(RasterizerTemporaryMemory& res_cache_tmp_mem, u32 mip_map,
+                               GLuint read_fb_handle, GLuint draw_fb_handle);
 
     void EnsureTextureDiscrepantView();
 
     OGLTexture texture;
     OGLTexture discrepant_view;
-    std::vector<std::vector<u8>> gl_buffer;
     SurfaceParams params{};
     GLenum gl_target{};
     GLenum gl_internal_format{};
@@ -473,6 +476,11 @@ public:
     void SignalPreDrawCall();
     void SignalPostDrawCall();
 
+protected:
+    void FlushObjectInner(const Surface& object) override {
+        object->FlushGLBuffer(temporal_memory);
+    }
+
 private:
     void LoadSurface(const Surface& surface);
     Surface GetSurface(const SurfaceParams& params, bool preserve_contents = true);
@@ -519,6 +527,8 @@ private:
     std::array<Surface, Maxwell::NumRenderTargets> current_color_buffers;
     Surface last_depth_buffer;
 
+    RasterizerTemporaryMemory temporal_memory;
+
     using SurfaceIntervalCache = boost::icl::interval_map<CacheAddr, Surface>;
     using SurfaceInterval = typename SurfaceIntervalCache::interval_type;
 
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index b1c8f7c35..f700dc89a 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -345,7 +345,7 @@ ShaderDiskCacheUsage CachedShader::GetUsage(GLenum primitive_mode,
 
 ShaderCacheOpenGL::ShaderCacheOpenGL(RasterizerOpenGL& rasterizer, Core::System& system,
                                      const Device& device)
-    : RasterizerCache{rasterizer}, disk_cache{system}, device{device} {}
+    : RasterizerCache{rasterizer}, device{device}, disk_cache{system} {}
 
 void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
                                       const VideoCore::DiskResourceLoadCallback& callback) {
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index a332087f8..31b979987 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -57,9 +57,6 @@ public:
         return shader_length;
     }
 
-    // We do not have to flush this cache as things in it are never modified by us.
-    void Flush() override {}
-
     /// Gets the shader entries for the shader
     const GLShader::ShaderEntries& GetShaderEntries() const {
         return entries;
@@ -123,6 +120,10 @@ public:
     /// Gets the current specified shader stage program
     Shader GetStageProgram(Maxwell::ShaderProgram program);
 
+protected:
+    // We do not have to flush this cache as things in it are never modified by us.
+    void FlushObjectInner(const Shader& object) override {}
+
 private:
     std::unordered_map<u64, UnspecializedShader> GenerateUnspecializedShaders(
         const std::atomic_bool& stop_loading, const VideoCore::DiskResourceLoadCallback& callback,
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index ef1a1995f..1a62795e1 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -871,17 +871,6 @@ private:
         return {};
     }
 
-    std::string Composite(Operation operation) {
-        std::string value = "vec4(";
-        for (std::size_t i = 0; i < 4; ++i) {
-            value += Visit(operation[i]);
-            if (i < 3)
-                value += ", ";
-        }
-        value += ')';
-        return value;
-    }
-
     template <Type type>
     std::string Add(Operation operation) {
         return GenerateBinaryInfix(operation, "+", type, type, type);
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 ed7afc4a0..254c0d499 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -475,7 +475,10 @@ void ShaderDiskCacheOpenGL::SaveUsage(const ShaderDiskCacheUsage& usage) {
     ASSERT_MSG(it != transferable.end(), "Saving shader usage without storing raw previously");
 
     auto& usages{it->second};
-    ASSERT(usages.find(usage) == usages.end());
+    if (usages.find(usage) != usages.end()) {
+        // Skip this variant since the shader is already stored.
+        return;
+    }
     usages.insert(usage);
 
     FileUtil::IOFile file = AppendTransferableFile();
diff --git a/src/video_core/renderer_vulkan/vk_buffer_cache.h b/src/video_core/renderer_vulkan/vk_buffer_cache.h
index 08b786aad..3edf460df 100644
--- a/src/video_core/renderer_vulkan/vk_buffer_cache.h
+++ b/src/video_core/renderer_vulkan/vk_buffer_cache.h
@@ -49,9 +49,6 @@ public:
         return alignment;
     }
 
-    // We do not have to flush this cache as things in it are never modified by us.
-    void Flush() override {}
-
 private:
     VAddr cpu_addr{};
     std::size_t size{};
@@ -87,6 +84,10 @@ public:
         return buffer_handle;
     }
 
+protected:
+    // We do not have to flush this cache as things in it are never modified by us.
+    void FlushObjectInner(const std::shared_ptr<CachedBufferEntry>& object) override {}
+
 private:
     void AlignBuffer(std::size_t alignment);
 
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 23d9b10db..a11000f6b 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -315,7 +315,6 @@ private:
         constexpr std::array<const char*, INTERNAL_FLAGS_COUNT> names = {"zero", "sign", "carry",
                                                                          "overflow"};
         for (std::size_t flag = 0; flag < INTERNAL_FLAGS_COUNT; ++flag) {
-            const auto flag_code = static_cast<InternalFlag>(flag);
             const Id id = OpVariable(t_prv_bool, spv::StorageClass::Private, v_false);
             internal_flags[flag] = AddGlobalVariable(Name(id, names[flag]));
         }
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 819cc6131..5b033126d 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -540,8 +540,6 @@ Node4 ShaderIR::GetTexsCode(Instruction instr, TextureType texture_type,
 Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool depth_compare,
                             bool is_array, bool is_aoffi) {
     const std::size_t coord_count = GetCoordCount(texture_type);
-    const std::size_t total_coord_count = coord_count + (is_array ? 1 : 0);
-    const std::size_t total_reg_count = total_coord_count + (depth_compare ? 1 : 0);
 
     // If enabled arrays index is always stored in the gpr8 field
     const u64 array_register = instr.gpr8.Value();
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp
index b508d64e9..a9b8f69af 100644
--- a/src/video_core/textures/astc.cpp
+++ b/src/video_core/textures/astc.cpp
@@ -25,8 +25,8 @@
 
 class InputBitStream {
 public:
-    explicit InputBitStream(const unsigned char* ptr, int nBits = 0, int start_offset = 0)
-        : m_NumBits(nBits), m_CurByte(ptr), m_NextBit(start_offset % 8) {}
+    explicit InputBitStream(const unsigned char* ptr, int start_offset = 0)
+        : m_CurByte(ptr), m_NextBit(start_offset % 8) {}
 
     ~InputBitStream() = default;
 
@@ -55,12 +55,9 @@ public:
     }
 
 private:
-    const int m_NumBits;
     const unsigned char* m_CurByte;
     int m_NextBit = 0;
     int m_BitsRead = 0;
-
-    bool done = false;
 };
 
 class OutputBitStream {
@@ -114,7 +111,6 @@ private:
     const int m_NumBits;
     unsigned char* m_CurByte;
     int m_NextBit = 0;
-    int m_BitsRead = 0;
 
     bool done = false;
 };
@@ -1616,6 +1612,7 @@ 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::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) {
@@ -1630,7 +1627,7 @@ std::vector<uint8_t> Decompress(const uint8_t* data, uint32_t width, uint32_t he
                 uint32_t decompWidth = std::min(block_width, width - i);
                 uint32_t decompHeight = std::min(block_height, height - j);
 
-                uint8_t* outRow = outData.data() + (j * width + i) * 4;
+                uint8_t* outRow = depth_offset + outData.data() + (j * width + i) * 4;
                 for (uint32_t jj = 0; jj < decompHeight; jj++) {
                     memcpy(outRow + jj * width * 4, uncompData + jj * block_width, decompWidth * 4);
                 }
@@ -1638,6 +1635,7 @@ std::vector<uint8_t> Decompress(const uint8_t* data, uint32_t width, uint32_t he
                 blockIdx++;
             }
         }
+        depth_offset += height * width * 4;
     }
 
     return outData;
diff --git a/src/yuzu/compatdb.cpp b/src/yuzu/compatdb.cpp
index c8b0a5ec0..5477f050c 100644
--- a/src/yuzu/compatdb.cpp
+++ b/src/yuzu/compatdb.cpp
@@ -58,7 +58,7 @@ void CompatDB::Submit() {
 
         button(NextButton)->setEnabled(false);
         button(NextButton)->setText(tr("Submitting"));
-        button(QWizard::CancelButton)->setVisible(false);
+        button(CancelButton)->setVisible(false);
 
         testcase_watcher.setFuture(QtConcurrent::run(
             [] { return Core::System::GetInstance().TelemetrySession().SubmitTestcase(); }));
@@ -74,12 +74,12 @@ void CompatDB::OnTestcaseSubmitted() {
                               tr("An error occured while sending the Testcase"));
         button(NextButton)->setEnabled(true);
         button(NextButton)->setText(tr("Next"));
-        button(QWizard::CancelButton)->setVisible(true);
+        button(CancelButton)->setVisible(true);
     } else {
         next();
         // older versions of QT don't support the "NoCancelButtonOnLastPage" option, this is a
         // workaround
-        button(QWizard::CancelButton)->setVisible(false);
+        button(CancelButton)->setVisible(false);
     }
 }
 
diff --git a/src/yuzu/configuration/configure_dialog.cpp b/src/yuzu/configuration/configure_dialog.cpp
index a5218b051..32c05b797 100644
--- a/src/yuzu/configuration/configure_dialog.cpp
+++ b/src/yuzu/configuration/configure_dialog.cpp
@@ -17,8 +17,12 @@ ConfigureDialog::ConfigureDialog(QWidget* parent, HotkeyRegistry& registry)
     ui->hotkeysTab->Populate(registry);
     this->setConfiguration();
     this->PopulateSelectionList();
+
+    setWindowFlags(windowFlags() & ~Qt::WindowContextHelpButtonHint);
+
     connect(ui->selectorList, &QListWidget::itemSelectionChanged, this,
             &ConfigureDialog::UpdateVisibleTabs);
+
     adjustSize();
     ui->selectorList->setCurrentRow(0);
 
diff --git a/src/yuzu/configuration/configure_graphics.cpp b/src/yuzu/configuration/configure_graphics.cpp
index c299c0b5b..08ea41b0f 100644
--- a/src/yuzu/configuration/configure_graphics.cpp
+++ b/src/yuzu/configuration/configure_graphics.cpp
@@ -69,16 +69,20 @@ ConfigureGraphics::ConfigureGraphics(QWidget* parent)
 ConfigureGraphics::~ConfigureGraphics() = default;
 
 void ConfigureGraphics::setConfiguration() {
+    const bool runtime_lock = !Core::System::GetInstance().IsPoweredOn();
+
     ui->resolution_factor_combobox->setCurrentIndex(
         static_cast<int>(FromResolutionFactor(Settings::values.resolution_factor)));
     ui->toggle_frame_limit->setChecked(Settings::values.use_frame_limit);
     ui->frame_limit->setValue(Settings::values.frame_limit);
+    ui->use_compatibility_profile->setEnabled(runtime_lock);
     ui->use_compatibility_profile->setChecked(Settings::values.use_compatibility_profile);
+    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(!Core::System::GetInstance().IsPoweredOn());
+    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(!Core::System::GetInstance().IsPoweredOn());
+    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));
diff --git a/src/yuzu/hotkeys.h b/src/yuzu/hotkeys.h
index 4f526dc7e..248fadaf3 100644
--- a/src/yuzu/hotkeys.h
+++ b/src/yuzu/hotkeys.h
@@ -67,8 +67,6 @@ public:
 
 private:
     struct Hotkey {
-        Hotkey() : shortcut(nullptr), context(Qt::WindowShortcut) {}
-
         QKeySequence keyseq;
         QShortcut* shortcut = nullptr;
         Qt::ShortcutContext context = Qt::WindowShortcut;
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
index 68a176032..8f104062d 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
@@ -176,9 +176,13 @@ EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
 
     SDL_SetMainReady();
 
+    const SDL_GLprofile profile = Settings::values.use_compatibility_profile
+                                      ? SDL_GL_CONTEXT_PROFILE_COMPATIBILITY
+                                      : SDL_GL_CONTEXT_PROFILE_CORE;
+
     SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 4);
     SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 3);
-    SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_CORE);
+    SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, profile);
     SDL_GL_SetAttribute(SDL_GL_DOUBLEBUFFER, 1);
     SDL_GL_SetAttribute(SDL_GL_RED_SIZE, 8);
     SDL_GL_SetAttribute(SDL_GL_GREEN_SIZE, 8);
diff --git a/src/yuzu_cmd/yuzu.cpp b/src/yuzu_cmd/yuzu.cpp
index a1d7879b1..d3734927b 100644
--- a/src/yuzu_cmd/yuzu.cpp
+++ b/src/yuzu_cmd/yuzu.cpp
@@ -222,6 +222,7 @@ int main(int argc, char** argv) {
 
     system.TelemetrySession().AddField(Telemetry::FieldType::App, "Frontend", "SDL");
 
+    emu_window->MakeCurrent();
     system.Renderer().Rasterizer().LoadDiskResources();
 
     while (emu_window->IsOpen()) {