55 files changed, 854 insertions, 444 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/build.sh b/.travis/linux-mingw/build.sh
index be03cc0f3..b12d70b12 100755
--- a/.travis/linux-mingw/build.sh
+++ b/.travis/linux-mingw/build.sh
@@ -1,3 +1,3 @@
 #!/bin/bash -ex
 mkdir "$HOME/.ccache" || true
-docker run --env-file .travis/common/travis-ci.env -v $(pwd):/yuzu -v "$HOME/.ccache":/root/.ccache ubuntu:18.04 /bin/bash -ex /yuzu/.travis/linux-mingw/docker.sh
+docker run --env-file .travis/common/travis-ci.env -v $(pwd):/yuzu -v "$HOME/.ccache":/root/.ccache yuzuemu/build-environments:linux-mingw /bin/bash -ex /yuzu/.travis/linux-mingw/docker.sh
diff --git a/.travis/linux-mingw/deps.sh b/.travis/linux-mingw/deps.sh
index 540bb934a..55b5d6006 100755
--- a/.travis/linux-mingw/deps.sh
+++ b/.travis/linux-mingw/deps.sh
@@ -1,3 +1,3 @@
 #!/bin/sh -ex
-docker pull ubuntu:18.04
+docker pull yuzuemu/build-environments:linux-mingw
diff --git a/.travis/linux-mingw/docker.sh b/.travis/linux-mingw/docker.sh
index 6cf43a006..28033acfb 100755
--- a/.travis/linux-mingw/docker.sh
+++ b/.travis/linux-mingw/docker.sh
@@ -1,16 +1,6 @@
 #!/bin/bash -ex
 cd /yuzu
-MINGW_PACKAGES="sdl2-mingw-w64 qt5base-mingw-w64 qt5tools-mingw-w64 libsamplerate-mingw-w64 qt5multimedia-mingw-w64"
-apt-get update
-apt-get install -y gpg wget git python3-pip python ccache g++-mingw-w64-x86-64 gcc-mingw-w64-x86-64 mingw-w64-tools cmake
-echo 'deb http://ppa.launchpad.net/tobydox/mingw-w64/ubuntu bionic main ' > /etc/apt/sources.list.d/extras.list
-apt-key adv --keyserver keyserver.ubuntu.com --recv '72931B477E22FEFD47F8DECE02FE5F12ADDE29B2'
-apt-get update
-apt-get install -y ${MINGW_PACKAGES}
-# fix a problem in current MinGW headers
-wget -q https://raw.githubusercontent.com/Alexpux/mingw-w64/d0d7f784833bbb0b2d279310ddc6afb52fe47a46/mingw-w64-headers/crt/errno.h -O /usr/x86_64-w64-mingw32/include/errno.h
 # override Travis CI unreasonable ccache size
 echo 'max_size = 3.0G' > "$HOME/.ccache/ccache.conf"
@@ -23,8 +13,8 @@ echo '' >> /bin/cmd
 chmod +x /bin/cmd
 mkdir build && cd build
-cmake .. -DCMAKE_TOOLCHAIN_FILE="$(pwd)/../CMakeModules/MinGWCross.cmake" -DUSE_CCACHE=ON -DYUZU_USE_BUNDLED_UNICORN=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DCMAKE_BUILD_TYPE=Release
+cmake .. -G Ninja -DCMAKE_TOOLCHAIN_FILE="$(pwd)/../CMakeModules/MinGWCross.cmake" -DUSE_CCACHE=ON -DYUZU_USE_BUNDLED_UNICORN=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DCMAKE_BUILD_TYPE=Release
-make -j4
+ninja
 # Clean up the dirty hacks
 rm /bin/uname && mv /bin/uname1 /bin/uname
diff --git a/.travis/linux/build.sh b/.travis/linux/build.sh
index 2fced727d..3929f97fc 100755
--- a/.travis/linux/build.sh
+++ b/.travis/linux/build.sh
@@ -1,4 +1,4 @@
 #!/bin/bash -ex
 mkdir -p "$HOME/.ccache"
-docker run -e ENABLE_COMPATIBILITY_REPORTING --env-file .travis/common/travis-ci.env -v $(pwd):/yuzu -v "$HOME/.ccache":/root/.ccache ubuntu:18.04 /bin/bash /yuzu/.travis/linux/docker.sh
+docker run -e ENABLE_COMPATIBILITY_REPORTING --env-file .travis/common/travis-ci.env -v $(pwd):/yuzu -v "$HOME/.ccache":/root/.ccache yuzuemu/build-environments:linux-fresh /bin/bash /yuzu/.travis/linux/docker.sh
diff --git a/.travis/linux/deps.sh b/.travis/linux/deps.sh
index 540bb934a..8d23c517d 100755
--- a/.travis/linux/deps.sh
+++ b/.travis/linux/deps.sh
@@ -1,3 +1,3 @@
 #!/bin/sh -ex
-docker pull ubuntu:18.04
+docker pull yuzuemu/build-environments:linux-fresh
diff --git a/.travis/linux/docker.sh b/.travis/linux/docker.sh
index 8b7e65911..3a9970384 100755
--- a/.travis/linux/docker.sh
+++ b/.travis/linux/docker.sh
@@ -1,12 +1,9 @@
 #!/bin/bash -ex
-apt-get update
-apt-get install --no-install-recommends -y build-essential git libqt5opengl5-dev libsdl2-dev libssl-dev python qtbase5-dev qtwebengine5-dev wget cmake ninja-build ccache
 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/CMakeModules/GenerateSCMRev.cmake b/CMakeModules/GenerateSCMRev.cmake
index 08315a1f1..5e00d839f 100644
--- a/CMakeModules/GenerateSCMRev.cmake
+++ b/CMakeModules/GenerateSCMRev.cmake
@@ -19,7 +19,7 @@ set(BUILD_VERSION "0")
 if (BUILD_REPOSITORY)
  # regex capture the string nightly or canary into CMAKE_MATCH_1
  string(REGEX MATCH "yuzu-emu/yuzu-?(.*)" OUTVAR ${BUILD_REPOSITORY})
-  if (${CMAKE_MATCH_COUNT} GREATER 0)
+  if ("${CMAKE_MATCH_COUNT}" GREATER 0)
    # capitalize the first letter of each word in the repo name.
    string(REPLACE "-" ";" REPO_NAME_LIST ${CMAKE_MATCH_1})
    foreach(WORD ${REPO_NAME_LIST})
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..a1d87bbbc 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -21,15 +21,27 @@ if (MSVC)
    # Ensure that projects build with Unicode support.
    add_definitions(-DUNICODE -D_UNICODE)
-    # /W3 - Level 3 warnings
+    # /W3                 - Level 3 warnings
-    # /MP - Multi-threaded compilation
+    # /MP                 - Multi-threaded compilation
-    # /Zi - Output debugging information
+    # /Zi                 - Output debugging information
-    # /Zo - enhanced debug info for optimized builds
+    # /Zo                 - Enhanced debug info for optimized builds
-    # /permissive- - enables stricter C++ standards conformance checks
+    # /permissive-        - Enables stricter C++ standards conformance checks
-    # /EHsc - C++-only exception handling semantics
+    # /EHsc               - C++-only exception handling semantics
-    # /Zc:throwingNew - let codegen assume `operator new` will never return null
+    # /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:inline          - Let codegen omit inline functions in object files
-    add_compile_options(/W3 /MP /Zi /Zo /permissive- /EHsc /std:c++latest /Zc:throwingNew,inline)
+    # /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
+        /Zc:externConstexpr
+        /Zc:inline
+        /Zc:throwingNew
+    )
    # /GS- - No stack buffer overflow checks
    add_compile_options("$<$<CONFIG:Release>:/GS->")
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/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index 20d01fc88..0775a89fb 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -241,7 +241,8 @@ void Process::LoadModule(CodeSet module_, VAddr base_addr) {
 }
 Process::Process(Core::System& system)
-    : WaitObject{system.Kernel()}, address_arbiter{system}, mutex{system}, system{system} {}
+    : WaitObject{system.Kernel()}, vm_manager{system},
+      address_arbiter{system}, mutex{system}, system{system} {}
 Process::~Process() = default;
diff --git a/src/core/hle/kernel/vm_manager.cpp b/src/core/hle/kernel/vm_manager.cpp
index f0c0c12fc..48b13cfdd 100644
--- a/src/core/hle/kernel/vm_manager.cpp
+++ b/src/core/hle/kernel/vm_manager.cpp
@@ -62,7 +62,7 @@ bool VirtualMemoryArea::CanBeMergedWith(const VirtualMemoryArea& next) const {
    return true;
 }
-VMManager::VMManager() {
+VMManager::VMManager(Core::System& system) : system{system} {
    // Default to assuming a 39-bit address space. This way we have a sane
    // starting point with executables that don't provide metadata.
    Reset(FileSys::ProgramAddressSpaceType::Is39Bit);
@@ -111,7 +111,6 @@ ResultVal<VMManager::VMAHandle> VMManager::MapMemoryBlock(VAddr target,
    VirtualMemoryArea& final_vma = vma_handle->second;
    ASSERT(final_vma.size == size);
-    auto& system = Core::System::GetInstance();
    system.ArmInterface(0).MapBackingMemory(target, size, block->data() + offset,
                                            VMAPermission::ReadWriteExecute);
    system.ArmInterface(1).MapBackingMemory(target, size, block->data() + offset,
@@ -140,7 +139,6 @@ ResultVal<VMManager::VMAHandle> VMManager::MapBackingMemory(VAddr target, u8* me
    VirtualMemoryArea& final_vma = vma_handle->second;
    ASSERT(final_vma.size == size);
-    auto& system = Core::System::GetInstance();
    system.ArmInterface(0).MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
    system.ArmInterface(1).MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
    system.ArmInterface(2).MapBackingMemory(target, size, memory, VMAPermission::ReadWriteExecute);
@@ -223,7 +221,6 @@ ResultCode VMManager::UnmapRange(VAddr target, u64 size) {
    ASSERT(FindVMA(target)->second.size >= size);
-    auto& system = Core::System::GetInstance();
    system.ArmInterface(0).UnmapMemory(target, size);
    system.ArmInterface(1).UnmapMemory(target, size);
    system.ArmInterface(2).UnmapMemory(target, size);
@@ -376,7 +373,7 @@ ResultCode VMManager::UnmapCodeMemory(VAddr dst_address, VAddr src_address, u64
    Reprotect(src_vma_iter, VMAPermission::ReadWrite);
    if (dst_memory_state == MemoryState::ModuleCode) {
-        Core::System::GetInstance().InvalidateCpuInstructionCaches();
+        system.InvalidateCpuInstructionCaches();
    }
    return unmap_result;
diff --git a/src/core/hle/kernel/vm_manager.h b/src/core/hle/kernel/vm_manager.h
index 288eb9450..ec84d9a70 100644
--- a/src/core/hle/kernel/vm_manager.h
+++ b/src/core/hle/kernel/vm_manager.h
@@ -14,6 +14,10 @@
 #include "core/hle/result.h"
 #include "core/memory.h"
+namespace Core {
+class System;
+}
 namespace FileSys {
 enum class ProgramAddressSpaceType : u8;
 }
@@ -321,7 +325,7 @@ class VMManager final {
 public:
    using VMAHandle = VMAMap::const_iterator;
-    VMManager();
+    explicit VMManager(Core::System& system);
    ~VMManager();
    /// Clears the address space map, re-initializing with a single free area.
@@ -712,5 +716,7 @@ private:
    // The end of the currently allocated heap. This is not an inclusive
    // end of the range. This is essentially 'base_address + current_size'.
    VAddr heap_end = 0;
+    Core::System& system;
 };
 } // namespace Kernel
diff --git a/src/core/loader/nso.cpp b/src/core/loader/nso.cpp
index a86653204..8592b1f44 100644
--- a/src/core/loader/nso.cpp
+++ b/src/core/loader/nso.cpp
@@ -21,8 +21,6 @@
 #include "core/memory.h"
 #include "core/settings.h"
-#pragma optimize("", off)
 namespace Loader {
 namespace {
 struct MODHeader {
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/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index 6674d9405..036e66f05 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -105,6 +105,8 @@ bool DmaPusher::Step() {
                dma_state.non_incrementing = false;
                dma_increment_once = true;
                break;
+            default:
+                break;
            }
        }
    }
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
+        ProcessLaunch();
-        // kernels)
-        UNREACHABLE_MSG("Compute shaders are not implemented");
        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,
+KeplerMemory::KeplerMemory(Core::System& system, MemoryManager& memory_manager)
-                           MemoryManager& memory_manager)
+    : system{system}, memory_manager{memory_manager}, upload_state{memory_manager, regs.upload} {}
-    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager} {}
 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,
+    KeplerMemory(Core::System& system, MemoryManager& memory_manager);
-                 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;
+                Upload::Registers upload;
-                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;
                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;
+    Upload::State upload_state;
-    void ProcessExec();
-    void ProcessData(u32 data, bool is_last_call);
 };
 #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(upload, 0x60);
-ASSERT_REG_POSITION(line_count, 0x61);
-ASSERT_REG_POSITION(dest, 0x62);
 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{
+    : system{system}, rasterizer{rasterizer}, memory_manager{memory_manager},
-                                                                                  *this} {
+      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 cc2424d38..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))
@@ -243,9 +250,10 @@ public:
                    return "10_10_10_2";
                case Size::Size_11_11_10:
                    return "11_11_10";
+                default:
+                    UNREACHABLE();
+                    return {};
                }
-                UNREACHABLE();
-                return {};
            }
            std::string TypeString() const {
@@ -579,7 +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 {
@@ -1175,6 +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;
@@ -1218,6 +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)};
+        const std::size_t dst_size = regs.dst_pitch * regs.y_count;
-    auto dst_ptr{memory_manager.GetPointer(dest)};
-    if (!source_ptr) {
+        if (read_buffer.size() < src_size) {
-        LOG_ERROR(HW_GPU, "source_ptr is invalid");
+            read_buffer.resize(src_size);
-        return;
+        }
-    }
-    if (!dst_ptr) {
+        if (write_buffer.size() < dst_size) {
-        LOG_ERROR(HW_GPU, "dst_ptr is invalid");
+            write_buffer.resize(dst_size);
-        return;
+        }
-    }
-    const auto FlushAndInvalidate = [&](u32 src_size, u64 dst_size) {
+        memory_manager.ReadBlock(source, read_buffer.data(), src_size);
-        // TODO(Subv): For now, manually flush the regions until we implement GPU-accelerated
+        memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
-        // copying.
-        rasterizer.FlushRegion(ToCacheAddr(source_ptr), src_size);
-        // We have to invalidate the destination region to evict any outdated surfaces from the
+        Texture::UnswizzleSubrect(regs.x_count, regs.y_count, regs.dst_pitch,
-        // cache. We do this before actually writing the new data because the destination address
+                                  regs.src_params.size_x, src_bytes_per_pixel, read_buffer.data(),
-        // might contain a dirty surface that will have to be written back to memory.
+                                  write_buffer.data(), regs.src_params.BlockHeight(),
-        rasterizer.InvalidateRegion(ToCacheAddr(dst_ptr), dst_size);
+                                  regs.src_params.pos_x, regs.src_params.pos_y);
-    };
-    if (regs.exec.is_dst_linear && !regs.exec.is_src_linear) {
+        memory_manager.WriteBlock(dest, write_buffer.data(), dst_size);
-        ASSERT(regs.src_params.size_z == 1);
+    } else {
-        // If the input is tiled and the output is linear, deswizzle the input and copy it over.
+        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,
+        const std::size_t dst_size = Texture::CalculateSize(
-                           copy_size * src_bytes_per_pixel);
+            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,
+        const std::size_t dst_layer_size = Texture::CalculateSize(
-                                  regs.src_params.size_x, src_bytes_per_pixel, source_ptr, dst_ptr,
+            true, src_bytes_per_pixel, regs.dst_params.size_x, regs.dst_params.size_y, 1,
-                                  regs.src_params.BlockHeight(), regs.src_params.pos_x,
+            regs.dst_params.BlockHeight(), regs.dst_params.BlockDepth());
-                                  regs.src_params.pos_y);
-    } else {
-        ASSERT(regs.dst_params.size_z == 1);
-        ASSERT(regs.src_pitch == regs.x_count);
-        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,
+        if (read_buffer.size() < src_size) {
-                           regs.dst_params.size_x * regs.dst_params.size_y * src_bpp);
+            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/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 9a088a503..3cc945235 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -305,6 +305,8 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
            case Maxwell::ShaderProgram::Geometry:
                shader_program_manager->UseTrivialGeometryShader();
                break;
+            default:
+                break;
            }
            continue;
        }
@@ -920,8 +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_far = src.depth_range_far;
-        viewport.depth_range_near = regs.viewports[i].depth_range_near;
+        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_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 2a81b1169..b1c8f7c35 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -363,6 +363,10 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
    if (stop_loading)
        return;
+    // Track if precompiled cache was altered during loading to know if we have to serialize the
+    // virtual precompiled cache file back to the hard drive
+    bool precompiled_cache_altered = false;
    // Build shaders
    if (callback)
        callback(VideoCore::LoadCallbackStage::Build, 0, usages.size());
@@ -384,6 +388,7 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
            if (!shader) {
                // Invalidate the precompiled cache if a shader dumped shader was rejected
                disk_cache.InvalidatePrecompiled();
+                precompiled_cache_altered = true;
                dumps.clear();
            }
        }
@@ -405,8 +410,13 @@ void ShaderCacheOpenGL::LoadDiskCache(const std::atomic_bool& stop_loading,
        if (dumps.find(usage) == dumps.end()) {
            const auto& program = precompiled_programs.at(usage);
            disk_cache.SaveDump(usage, program->handle);
+            precompiled_cache_altered = true;
        }
    }
+    if (precompiled_cache_altered) {
+        disk_cache.SaveVirtualPrecompiledFile();
+    }
 }
 CachedProgram ShaderCacheOpenGL::GeneratePrecompiledProgram(
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 53752b38d..254c0d499 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.cpp
@@ -104,7 +104,8 @@ bool ShaderDiskCacheRaw::Save(FileUtil::IOFile& file) const {
    return true;
 }
-ShaderDiskCacheOpenGL::ShaderDiskCacheOpenGL(Core::System& system) : system{system} {}
+ShaderDiskCacheOpenGL::ShaderDiskCacheOpenGL(Core::System& system)
+    : system{system}, precompiled_cache_virtual_file_offset{0} {}
 std::optional<std::pair<std::vector<ShaderDiskCacheRaw>, std::vector<ShaderDiskCacheUsage>>>
 ShaderDiskCacheOpenGL::LoadTransferable() {
@@ -177,6 +178,7 @@ ShaderDiskCacheOpenGL::LoadTransferable() {
            return {};
        }
    }
    return {{raws, usages}};
 }
@@ -208,59 +210,64 @@ ShaderDiskCacheOpenGL::LoadPrecompiled() {
 std::optional<std::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>,
                        std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>>
 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());
+    const std::vector<u8> decompressed = Common::Compression::DecompressDataZSTD(compressed);
+    SaveArrayToPrecompiled(decompressed.data(), decompressed.size());
+    precompiled_cache_virtual_file_offset = 0;
    ShaderCacheVersionHash file_hash{};
-    if (file.ReadArray(file_hash.data(), file_hash.size()) != file_hash.size()) {
+    if (!LoadArrayFromPrecompiled(file_hash.data(), file_hash.size())) {
+        precompiled_cache_virtual_file_offset = 0;
        return {};
    }
    if (GetShaderCacheVersionHash() != file_hash) {
        LOG_INFO(Render_OpenGL, "Precompiled cache is from another version of the emulator");
+        precompiled_cache_virtual_file_offset = 0;
        return {};
    }
    std::unordered_map<u64, ShaderDiskCacheDecompiled> decompiled;
    std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump> dumps;
-    while (file.Tell() < file.GetSize()) {
+    while (precompiled_cache_virtual_file_offset < precompiled_cache_virtual_file.GetSize()) {
        PrecompiledEntryKind kind{};
-        if (file.ReadBytes(&kind, sizeof(u32)) != sizeof(u32)) {
+        if (!LoadObjectFromPrecompiled(kind)) {
            return {};
        }
        switch (kind) {
        case PrecompiledEntryKind::Decompiled: {
            u64 unique_identifier{};
-            if (file.ReadBytes(&unique_identifier, sizeof(u64)) != sizeof(u64))
+            if (!LoadObjectFromPrecompiled(unique_identifier)) {
                return {};
+            }
-            const auto entry = LoadDecompiledEntry(file);
+            const auto entry = LoadDecompiledEntry();
-            if (!entry)
+            if (!entry) {
                return {};
+            }
            decompiled.insert({unique_identifier, std::move(*entry)});
            break;
        }
        case PrecompiledEntryKind::Dump: {
            ShaderDiskCacheUsage usage;
-            if (file.ReadBytes(&usage, sizeof(usage)) != sizeof(usage))
+            if (!LoadObjectFromPrecompiled(usage)) {
                return {};
+            }
            ShaderDiskCacheDump dump;
-            if (file.ReadBytes(&dump.binary_format, sizeof(u32)) != sizeof(u32))
+            if (!LoadObjectFromPrecompiled(dump.binary_format)) {
-                return {};
-            u32 binary_length{};
-            u32 compressed_size{};
-            if (file.ReadBytes(&binary_length, sizeof(u32)) != sizeof(u32) ||
-                file.ReadBytes(&compressed_size, sizeof(u32)) != sizeof(u32)) {
                return {};
            }
-            std::vector<u8> compressed_binary(compressed_size);
+            u32 binary_length{};
-            if (file.ReadArray(compressed_binary.data(), compressed_binary.size()) !=
+            if (!LoadObjectFromPrecompiled(binary_length)) {
-                compressed_binary.size()) {
                return {};
            }
-            dump.binary = Common::Compression::DecompressDataZSTD(compressed_binary);
+            dump.binary.resize(binary_length);
-            if (dump.binary.empty()) {
+            if (!LoadArrayFromPrecompiled(dump.binary.data(), dump.binary.size())) {
                return {};
            }
@@ -274,45 +281,41 @@ ShaderDiskCacheOpenGL::LoadPrecompiledFile(FileUtil::IOFile& file) {
    return {{decompiled, dumps}};
 }
-std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEntry(
+std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEntry() {
-    FileUtil::IOFile& file) {
    u32 code_size{};
-    u32 compressed_code_size{};
+    if (!LoadObjectFromPrecompiled(code_size)) {
-    if (file.ReadBytes(&code_size, sizeof(u32)) != sizeof(u32) ||
-        file.ReadBytes(&compressed_code_size, sizeof(u32)) != sizeof(u32)) {
        return {};
    }
-    std::vector<u8> compressed_code(compressed_code_size);
+    std::vector<u8> code(code_size);
-    if (file.ReadArray(compressed_code.data(), compressed_code.size()) != compressed_code.size()) {
+    if (!LoadArrayFromPrecompiled(code.data(), code.size())) {
        return {};
    }
-    const std::vector<u8> code = Common::Compression::DecompressDataZSTD(compressed_code);
-    if (code.empty()) {
-        return {};
-    }
    ShaderDiskCacheDecompiled entry;
    entry.code = std::string(reinterpret_cast<const char*>(code.data()), code_size);
    u32 const_buffers_count{};
-    if (file.ReadBytes(&const_buffers_count, sizeof(u32)) != sizeof(u32))
+    if (!LoadObjectFromPrecompiled(const_buffers_count)) {
        return {};
+    }
    for (u32 i = 0; i < const_buffers_count; ++i) {
        u32 max_offset{};
        u32 index{};
        u8 is_indirect{};
-        if (file.ReadBytes(&max_offset, sizeof(u32)) != sizeof(u32) ||
+        if (!LoadObjectFromPrecompiled(max_offset) || !LoadObjectFromPrecompiled(index) ||
-            file.ReadBytes(&index, sizeof(u32)) != sizeof(u32) ||
+            !LoadObjectFromPrecompiled(is_indirect)) {
-            file.ReadBytes(&is_indirect, sizeof(u8)) != sizeof(u8)) {
            return {};
        }
        entry.entries.const_buffers.emplace_back(max_offset, is_indirect != 0, index);
    }
    u32 samplers_count{};
-    if (file.ReadBytes(&samplers_count, sizeof(u32)) != sizeof(u32))
+    if (!LoadObjectFromPrecompiled(samplers_count)) {
        return {};
+    }
    for (u32 i = 0; i < samplers_count; ++i) {
        u64 offset{};
        u64 index{};
@@ -320,12 +323,9 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
        u8 is_array{};
        u8 is_shadow{};
        u8 is_bindless{};
-        if (file.ReadBytes(&offset, sizeof(u64)) != sizeof(u64) ||
+        if (!LoadObjectFromPrecompiled(offset) || !LoadObjectFromPrecompiled(index) ||
-            file.ReadBytes(&index, sizeof(u64)) != sizeof(u64) ||
+            !LoadObjectFromPrecompiled(type) || !LoadObjectFromPrecompiled(is_array) ||
-            file.ReadBytes(&type, sizeof(u32)) != sizeof(u32) ||
+            !LoadObjectFromPrecompiled(is_shadow) || !LoadObjectFromPrecompiled(is_bindless)) {
-            file.ReadBytes(&is_array, sizeof(u8)) != sizeof(u8) ||
-            file.ReadBytes(&is_shadow, sizeof(u8)) != sizeof(u8) ||
-            file.ReadBytes(&is_bindless, sizeof(u8)) != sizeof(u8)) {
            return {};
        }
        entry.entries.samplers.emplace_back(static_cast<std::size_t>(offset),
@@ -335,17 +335,17 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
    }
    u32 global_memory_count{};
-    if (file.ReadBytes(&global_memory_count, sizeof(u32)) != sizeof(u32))
+    if (!LoadObjectFromPrecompiled(global_memory_count)) {
        return {};
+    }
    for (u32 i = 0; i < global_memory_count; ++i) {
        u32 cbuf_index{};
        u32 cbuf_offset{};
        u8 is_read{};
        u8 is_written{};
-        if (file.ReadBytes(&cbuf_index, sizeof(u32)) != sizeof(u32) ||
+        if (!LoadObjectFromPrecompiled(cbuf_index) || !LoadObjectFromPrecompiled(cbuf_offset) ||
-            file.ReadBytes(&cbuf_offset, sizeof(u32)) != sizeof(u32) ||
+            !LoadObjectFromPrecompiled(is_read) || !LoadObjectFromPrecompiled(is_written)) {
-            file.ReadBytes(&is_read, sizeof(u8)) != sizeof(u8) ||
-            file.ReadBytes(&is_written, sizeof(u8)) != sizeof(u8)) {
            return {};
        }
        entry.entries.global_memory_entries.emplace_back(cbuf_index, cbuf_offset, is_read != 0,
@@ -354,74 +354,81 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
    for (auto& clip_distance : entry.entries.clip_distances) {
        u8 clip_distance_raw{};
-        if (file.ReadBytes(&clip_distance_raw, sizeof(u8)) != sizeof(u8))
+        if (!LoadObjectFromPrecompiled(clip_distance_raw))
            return {};
        clip_distance = clip_distance_raw != 0;
    }
    u64 shader_length{};
-    if (file.ReadBytes(&shader_length, sizeof(u64)) != sizeof(u64))
+    if (!LoadObjectFromPrecompiled(shader_length)) {
        return {};
+    }
    entry.entries.shader_length = static_cast<std::size_t>(shader_length);
    return entry;
 }
-bool ShaderDiskCacheOpenGL::SaveDecompiledFile(FileUtil::IOFile& file, u64 unique_identifier,
+bool ShaderDiskCacheOpenGL::SaveDecompiledFile(u64 unique_identifier, const std::string& code,
-                                               const std::string& code,
-                                               const std::vector<u8>& compressed_code,
                                               const GLShader::ShaderEntries& entries) {
-    if (file.WriteObject(static_cast<u32>(PrecompiledEntryKind::Decompiled)) != 1 ||
+    if (!SaveObjectToPrecompiled(static_cast<u32>(PrecompiledEntryKind::Decompiled)) ||
-        file.WriteObject(unique_identifier) != 1 ||
+        !SaveObjectToPrecompiled(unique_identifier) ||
-        file.WriteObject(static_cast<u32>(code.size())) != 1 ||
+        !SaveObjectToPrecompiled(static_cast<u32>(code.size())) ||
-        file.WriteObject(static_cast<u32>(compressed_code.size())) != 1 ||
+        !SaveArrayToPrecompiled(code.data(), code.size())) {
-        file.WriteArray(compressed_code.data(), compressed_code.size()) != compressed_code.size()) {
        return false;
    }
-    if (file.WriteObject(static_cast<u32>(entries.const_buffers.size())) != 1)
+    if (!SaveObjectToPrecompiled(static_cast<u32>(entries.const_buffers.size()))) {
        return false;
+    }
    for (const auto& cbuf : entries.const_buffers) {
-        if (file.WriteObject(static_cast<u32>(cbuf.GetMaxOffset())) != 1 ||
+        if (!SaveObjectToPrecompiled(static_cast<u32>(cbuf.GetMaxOffset())) ||
-            file.WriteObject(static_cast<u32>(cbuf.GetIndex())) != 1 ||
+            !SaveObjectToPrecompiled(static_cast<u32>(cbuf.GetIndex())) ||
-            file.WriteObject(static_cast<u8>(cbuf.IsIndirect() ? 1 : 0)) != 1) {
+            !SaveObjectToPrecompiled(static_cast<u8>(cbuf.IsIndirect() ? 1 : 0))) {
            return false;
        }
    }
-    if (file.WriteObject(static_cast<u32>(entries.samplers.size())) != 1)
+    if (!SaveObjectToPrecompiled(static_cast<u32>(entries.samplers.size()))) {
        return false;
+    }
    for (const auto& sampler : entries.samplers) {
-        if (file.WriteObject(static_cast<u64>(sampler.GetOffset())) != 1 ||
+        if (!SaveObjectToPrecompiled(static_cast<u64>(sampler.GetOffset())) ||
-            file.WriteObject(static_cast<u64>(sampler.GetIndex())) != 1 ||
+            !SaveObjectToPrecompiled(static_cast<u64>(sampler.GetIndex())) ||
-            file.WriteObject(static_cast<u32>(sampler.GetType())) != 1 ||
+            !SaveObjectToPrecompiled(static_cast<u32>(sampler.GetType())) ||
-            file.WriteObject(static_cast<u8>(sampler.IsArray() ? 1 : 0)) != 1 ||
+            !SaveObjectToPrecompiled(static_cast<u8>(sampler.IsArray() ? 1 : 0)) ||
-            file.WriteObject(static_cast<u8>(sampler.IsShadow() ? 1 : 0)) != 1 ||
+            !SaveObjectToPrecompiled(static_cast<u8>(sampler.IsShadow() ? 1 : 0)) ||
-            file.WriteObject(static_cast<u8>(sampler.IsBindless() ? 1 : 0)) != 1) {
+            !SaveObjectToPrecompiled(static_cast<u8>(sampler.IsBindless() ? 1 : 0))) {
            return false;
        }
    }
-    if (file.WriteObject(static_cast<u32>(entries.global_memory_entries.size())) != 1)
+    if (!SaveObjectToPrecompiled(static_cast<u32>(entries.global_memory_entries.size()))) {
        return false;
+    }
    for (const auto& gmem : entries.global_memory_entries) {
-        if (file.WriteObject(static_cast<u32>(gmem.GetCbufIndex())) != 1 ||
+        if (!SaveObjectToPrecompiled(static_cast<u32>(gmem.GetCbufIndex())) ||
-            file.WriteObject(static_cast<u32>(gmem.GetCbufOffset())) != 1 ||
+            !SaveObjectToPrecompiled(static_cast<u32>(gmem.GetCbufOffset())) ||
-            file.WriteObject(static_cast<u8>(gmem.IsRead() ? 1 : 0)) != 1 ||
+            !SaveObjectToPrecompiled(static_cast<u8>(gmem.IsRead() ? 1 : 0)) ||
-            file.WriteObject(static_cast<u8>(gmem.IsWritten() ? 1 : 0)) != 1) {
+            !SaveObjectToPrecompiled(static_cast<u8>(gmem.IsWritten() ? 1 : 0))) {
            return false;
        }
    }
    for (const bool clip_distance : entries.clip_distances) {
-        if (file.WriteObject(static_cast<u8>(clip_distance ? 1 : 0)) != 1)
+        if (!SaveObjectToPrecompiled(static_cast<u8>(clip_distance ? 1 : 0))) {
            return false;
+        }
    }
-    return file.WriteObject(static_cast<u64>(entries.shader_length)) == 1;
+    if (!SaveObjectToPrecompiled(static_cast<u64>(entries.shader_length))) {
+        return false;
+    }
+    return true;
 }
-void ShaderDiskCacheOpenGL::InvalidateTransferable() const {
+void ShaderDiskCacheOpenGL::InvalidateTransferable() {
    if (!FileUtil::Delete(GetTransferablePath())) {
        LOG_ERROR(Render_OpenGL, "Failed to invalidate transferable file={}",
                  GetTransferablePath());
@@ -429,7 +436,10 @@ void ShaderDiskCacheOpenGL::InvalidateTransferable() const {
    InvalidatePrecompiled();
 }
-void ShaderDiskCacheOpenGL::InvalidatePrecompiled() const {
+void ShaderDiskCacheOpenGL::InvalidatePrecompiled() {
+    // Clear virtaul precompiled cache file
+    precompiled_cache_virtual_file.Resize(0);
    if (!FileUtil::Delete(GetPrecompiledPath())) {
        LOG_ERROR(Render_OpenGL, "Failed to invalidate precompiled file={}", GetPrecompiledPath());
    }
@@ -465,7 +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();
@@ -485,22 +498,13 @@ void ShaderDiskCacheOpenGL::SaveDecompiled(u64 unique_identifier, const std::str
    if (!IsUsable())
        return;
-    const std::vector<u8> compressed_code{Common::Compression::CompressDataZSTDDefault(
+    if (precompiled_cache_virtual_file.GetSize() == 0) {
-        reinterpret_cast<const u8*>(code.data()), code.size())};
+        SavePrecompiledHeaderToVirtualPrecompiledCache();
-    if (compressed_code.empty()) {
-        LOG_ERROR(Render_OpenGL, "Failed to compress GLSL code - skipping shader {:016x}",
-                  unique_identifier);
-        return;
    }
-    FileUtil::IOFile file = AppendPrecompiledFile();
+    if (!SaveDecompiledFile(unique_identifier, code, entries)) {
-    if (!file.IsOpen())
-        return;
-    if (!SaveDecompiledFile(file, unique_identifier, code, compressed_code, entries)) {
        LOG_ERROR(Render_OpenGL,
                  "Failed to save decompiled entry to the precompiled file - removing");
-        file.Close();
        InvalidatePrecompiled();
    }
 }
@@ -516,28 +520,13 @@ void ShaderDiskCacheOpenGL::SaveDump(const ShaderDiskCacheUsage& usage, GLuint p
    std::vector<u8> binary(binary_length);
    glGetProgramBinary(program, binary_length, nullptr, &binary_format, binary.data());
-    const std::vector<u8> compressed_binary =
+    if (!SaveObjectToPrecompiled(static_cast<u32>(PrecompiledEntryKind::Dump)) ||
-        Common::Compression::CompressDataZSTDDefault(binary.data(), binary.size());
+        !SaveObjectToPrecompiled(usage) ||
+        !SaveObjectToPrecompiled(static_cast<u32>(binary_format)) ||
-    if (compressed_binary.empty()) {
+        !SaveObjectToPrecompiled(static_cast<u32>(binary_length)) ||
-        LOG_ERROR(Render_OpenGL, "Failed to compress binary program in shader={:016x}",
+        !SaveArrayToPrecompiled(binary.data(), binary.size())) {
-                  usage.unique_identifier);
-        return;
-    }
-    FileUtil::IOFile file = AppendPrecompiledFile();
-    if (!file.IsOpen())
-        return;
-    if (file.WriteObject(static_cast<u32>(PrecompiledEntryKind::Dump)) != 1 ||
-        file.WriteObject(usage) != 1 || file.WriteObject(static_cast<u32>(binary_format)) != 1 ||
-        file.WriteObject(static_cast<u32>(binary_length)) != 1 ||
-        file.WriteObject(static_cast<u32>(compressed_binary.size())) != 1 ||
-        file.WriteArray(compressed_binary.data(), compressed_binary.size()) !=
-            compressed_binary.size()) {
        LOG_ERROR(Render_OpenGL, "Failed to save binary program file in shader={:016x} - removing",
                  usage.unique_identifier);
-        file.Close();
        InvalidatePrecompiled();
        return;
    }
@@ -570,28 +559,33 @@ FileUtil::IOFile ShaderDiskCacheOpenGL::AppendTransferableFile() const {
    return file;
 }
-FileUtil::IOFile ShaderDiskCacheOpenGL::AppendPrecompiledFile() const {
+void ShaderDiskCacheOpenGL::SavePrecompiledHeaderToVirtualPrecompiledCache() {
-    if (!EnsureDirectories())
+    const auto hash{GetShaderCacheVersionHash()};
-        return {};
+    if (!SaveArrayToPrecompiled(hash.data(), hash.size())) {
+        LOG_ERROR(
+            Render_OpenGL,
+            "Failed to write precompiled cache version hash to virtual precompiled cache file");
+    }
+}
+void ShaderDiskCacheOpenGL::SaveVirtualPrecompiledFile() {
+    precompiled_cache_virtual_file_offset = 0;
+    const std::vector<u8>& uncompressed = precompiled_cache_virtual_file.ReadAllBytes();
+    const std::vector<u8>& compressed =
+        Common::Compression::CompressDataZSTDDefault(uncompressed.data(), uncompressed.size());
    const auto precompiled_path{GetPrecompiledPath()};
-    const bool existed = FileUtil::Exists(precompiled_path);
+    FileUtil::IOFile file(precompiled_path, "wb");
-    FileUtil::IOFile file(precompiled_path, "ab");
    if (!file.IsOpen()) {
        LOG_ERROR(Render_OpenGL, "Failed to open precompiled cache in path={}", precompiled_path);
-        return {};
+        return;
    }
+    if (file.WriteBytes(compressed.data(), compressed.size()) != compressed.size()) {
-    if (!existed || file.GetSize() == 0) {
+        LOG_ERROR(Render_OpenGL, "Failed to write precompiled cache version in path={}",
-        const auto hash{GetShaderCacheVersionHash()};
+                  precompiled_path);
-        if (file.WriteArray(hash.data(), hash.size()) != hash.size()) {
+        return;
-            LOG_ERROR(Render_OpenGL, "Failed to write precompiled cache version hash in path={}",
-                      precompiled_path);
-            return {};
-        }
    }
-    return file;
 }
 bool ShaderDiskCacheOpenGL::EnsureDirectories() const {
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 6be0c0547..0142b2e3b 100644
--- a/src/video_core/renderer_opengl/gl_shader_disk_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_disk_cache.h
@@ -16,6 +16,7 @@
 #include "common/assert.h"
 #include "common/common_types.h"
+#include "core/file_sys/vfs_vector.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
@@ -172,10 +173,10 @@ public:
    LoadPrecompiled();
    /// Removes the transferable (and precompiled) cache file.
-    void InvalidateTransferable() const;
+    void InvalidateTransferable();
-    /// Removes the precompiled cache file.
+    /// Removes the precompiled cache file and clears virtual precompiled cache file.
-    void InvalidatePrecompiled() const;
+    void InvalidatePrecompiled();
    /// Saves a raw dump to the transferable file. Checks for collisions.
    void SaveRaw(const ShaderDiskCacheRaw& entry);
@@ -190,18 +191,21 @@ public:
    /// Saves a dump entry to the precompiled file. Does not check for collisions.
    void SaveDump(const ShaderDiskCacheUsage& usage, 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::pair<std::unordered_map<u64, ShaderDiskCacheDecompiled>,
                            std::unordered_map<ShaderDiskCacheUsage, ShaderDiskCacheDump>>>
    LoadPrecompiledFile(FileUtil::IOFile& file);
-    /// Loads a decompiled cache entry from the passed file. Returns empty on failure.
+    /// Loads a decompiled cache entry from m_precompiled_cache_virtual_file. Returns empty on
-    std::optional<ShaderDiskCacheDecompiled> LoadDecompiledEntry(FileUtil::IOFile& file);
+    /// failure.
+    std::optional<ShaderDiskCacheDecompiled> LoadDecompiledEntry();
    /// Saves a decompiled entry to the passed file. Returns true on success.
-    bool SaveDecompiledFile(FileUtil::IOFile& file, u64 unique_identifier, const std::string& code,
+    bool SaveDecompiledFile(u64 unique_identifier, const std::string& code,
-                            const std::vector<u8>& compressed_code,
                            const GLShader::ShaderEntries& entries);
    /// Returns if the cache can be used
@@ -210,8 +214,8 @@ private:
    /// Opens current game's transferable file and write it's header if it doesn't exist
    FileUtil::IOFile AppendTransferableFile() const;
-    /// Opens current game's precompiled file and write it's header if it doesn't exist
+    /// Save precompiled header to precompiled_cache_in_memory
-    FileUtil::IOFile AppendPrecompiledFile() const;
+    void SavePrecompiledHeaderToVirtualPrecompiledCache();
    /// Create shader disk cache directories. Returns true on success.
    bool EnsureDirectories() const;
@@ -234,10 +238,42 @@ private:
    /// Get current game's title id
    std::string GetTitleID() const;
+    template <typename T>
+    bool SaveArrayToPrecompiled(const T* data, std::size_t length) {
+        const std::size_t write_length = precompiled_cache_virtual_file.WriteArray(
+            data, length, precompiled_cache_virtual_file_offset);
+        precompiled_cache_virtual_file_offset += write_length;
+        return write_length == sizeof(T) * length;
+    }
+    template <typename T>
+    bool LoadArrayFromPrecompiled(T* data, std::size_t length) {
+        const std::size_t read_length = precompiled_cache_virtual_file.ReadArray(
+            data, length, precompiled_cache_virtual_file_offset);
+        precompiled_cache_virtual_file_offset += read_length;
+        return read_length == sizeof(T) * length;
+    }
+    template <typename T>
+    bool SaveObjectToPrecompiled(const T& object) {
+        return SaveArrayToPrecompiled(&object, 1);
+    }
+    template <typename T>
+    bool LoadObjectFromPrecompiled(T& object) {
+        return LoadArrayFromPrecompiled(&object, 1);
+    }
    // Copre system
    Core::System& system;
    // Stored transferable shaders
    std::map<u64, std::unordered_set<ShaderDiskCacheUsage>> transferable;
+    // Stores whole precompiled cache which will be read from or saved to the precompiled chache
+    // file
+    FileSys::VectorVfsFile precompiled_cache_virtual_file;
+    // Stores the current offset of the precompiled cache file for IO purposes
+    std::size_t precompiled_cache_virtual_file_offset;
    // The cache has been loaded at boot
    bool tried_to_load{};
 };
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index a8833c06e..95b773135 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -27,8 +27,7 @@ using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
    switch (attrib.type) {
    case Maxwell::VertexAttribute::Type::UnsignedInt:
-    case Maxwell::VertexAttribute::Type::UnsignedNorm: {
+    case Maxwell::VertexAttribute::Type::UnsignedNorm:
        switch (attrib.size) {
        case Maxwell::VertexAttribute::Size::Size_8:
        case Maxwell::VertexAttribute::Size::Size_8_8:
@@ -47,16 +46,13 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
            return GL_UNSIGNED_INT;
        case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
            return GL_UNSIGNED_INT_2_10_10_10_REV;
+        default:
+            LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
+            UNREACHABLE();
+            return {};
        }
-        LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
-        UNREACHABLE();
-        return {};
-    }
    case Maxwell::VertexAttribute::Type::SignedInt:
-    case Maxwell::VertexAttribute::Type::SignedNorm: {
+    case Maxwell::VertexAttribute::Type::SignedNorm:
        switch (attrib.size) {
        case Maxwell::VertexAttribute::Size::Size_8:
        case Maxwell::VertexAttribute::Size::Size_8_8:
@@ -75,14 +71,12 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
            return GL_INT;
        case Maxwell::VertexAttribute::Size::Size_10_10_10_2:
            return GL_INT_2_10_10_10_REV;
+        default:
+            LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
+            UNREACHABLE();
+            return {};
        }
+    case Maxwell::VertexAttribute::Type::Float:
-        LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
-        UNREACHABLE();
-        return {};
-    }
-    case Maxwell::VertexAttribute::Type::Float: {
        switch (attrib.size) {
        case Maxwell::VertexAttribute::Size::Size_16:
        case Maxwell::VertexAttribute::Size::Size_16_16:
@@ -94,13 +88,16 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
        case Maxwell::VertexAttribute::Size::Size_32_32_32:
        case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
            return GL_FLOAT;
+        default:
+            LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
+            UNREACHABLE();
+            return {};
        }
+    default:
+        LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex type={}", attrib.TypeString());
+        UNREACHABLE();
+        return {};
    }
-    }
-    LOG_CRITICAL(Render_OpenGL, "Unimplemented vertex type={}", attrib.TypeString());
-    UNREACHABLE();
-    return {};
 }
 inline GLenum IndexFormat(Maxwell::IndexFormat index_format) {
@@ -129,10 +126,11 @@ inline GLenum PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
        return GL_TRIANGLES;
    case Maxwell::PrimitiveTopology::TriangleStrip:
        return GL_TRIANGLE_STRIP;
+    default:
+        LOG_CRITICAL(Render_OpenGL, "Unimplemented topology={}", static_cast<u32>(topology));
+        UNREACHABLE();
+        return {};
    }
-    LOG_CRITICAL(Render_OpenGL, "Unimplemented topology={}", static_cast<u32>(topology));
-    UNREACHABLE();
-    return {};
 }
 inline GLenum TextureFilterMode(Tegra::Texture::TextureFilter filter_mode,
@@ -186,9 +184,10 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) {
        } else {
            return GL_MIRROR_CLAMP_TO_EDGE;
        }
+    default:
+        LOG_ERROR(Render_OpenGL, "Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode));
+        return GL_REPEAT;
    }
-    LOG_ERROR(Render_OpenGL, "Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode));
-    return GL_REPEAT;
 }
 inline GLenum DepthCompareFunc(Tegra::Texture::DepthCompareFunc func) {
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 34bf26ff2..9fe1e3280 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -62,9 +62,10 @@ vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode) {
    case Tegra::Texture::WrapMode::MirrorOnceBorder:
        UNIMPLEMENTED();
        return vk::SamplerAddressMode::eMirrorClampToEdge;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode));
+        return {};
    }
-    UNIMPLEMENTED_MSG("Unimplemented wrap mode={}", static_cast<u32>(wrap_mode));
-    return {};
 }
 vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func) {
@@ -225,9 +226,10 @@ vk::PrimitiveTopology PrimitiveTopology(Maxwell::PrimitiveTopology topology) {
        return vk::PrimitiveTopology::eTriangleList;
    case Maxwell::PrimitiveTopology::TriangleStrip:
        return vk::PrimitiveTopology::eTriangleStrip;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology));
+        return {};
    }
-    UNIMPLEMENTED_MSG("Unimplemented topology={}", static_cast<u32>(topology));
-    return {};
 }
 vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttribute::Size size) {
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp
index e4c438792..2da595c0d 100644
--- a/src/video_core/shader/decode.cpp
+++ b/src/video_core/shader/decode.cpp
@@ -116,6 +116,8 @@ ExitMethod ShaderIR::Scan(u32 begin, u32 end, std::set<u32>& labels) {
            // Continue scanning for an exit method.
            break;
        }
+        default:
+            break;
        }
    }
    return exit_method = ExitMethod::AlwaysReturn;
@@ -206,4 +208,4 @@ u32 ShaderIR::DecodeInstr(NodeBlock& bb, u32 pc) {
    return pc + 1;
 }
-} // namespace VideoCommon::Shader
-\ No newline at end of file
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/arithmetic_half.cpp b/src/video_core/shader/decode/arithmetic_half.cpp
index 9467f9417..2098c1170 100644
--- a/src/video_core/shader/decode/arithmetic_half.cpp
+++ b/src/video_core/shader/decode/arithmetic_half.cpp
@@ -9,6 +9,7 @@
 namespace VideoCommon::Shader {
+using Tegra::Shader::HalfType;
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
@@ -22,7 +23,6 @@ u32 ShaderIR::DecodeArithmeticHalf(NodeBlock& bb, u32 pc) {
            LOG_WARNING(HW_GPU, "{} FTZ not implemented", opcode->get().GetName());
        }
    }
-    UNIMPLEMENTED_IF_MSG(instr.alu_half.saturate != 0, "Half float saturation not implemented");
    const bool negate_a =
        opcode->get().GetId() != OpCode::Id::HMUL2_R && instr.alu_half.negate_a != 0;
@@ -32,35 +32,37 @@ u32 ShaderIR::DecodeArithmeticHalf(NodeBlock& bb, u32 pc) {
    Node op_a = UnpackHalfFloat(GetRegister(instr.gpr8), instr.alu_half.type_a);
    op_a = GetOperandAbsNegHalf(op_a, instr.alu_half.abs_a, negate_a);
-    Node op_b = [&]() {
+    auto [type_b, op_b] = [&]() -> std::tuple<HalfType, Node> {
        switch (opcode->get().GetId()) {
        case OpCode::Id::HADD2_C:
        case OpCode::Id::HMUL2_C:
-            return GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset());
+            return {HalfType::F32, GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset())};
        case OpCode::Id::HADD2_R:
        case OpCode::Id::HMUL2_R:
-            return GetRegister(instr.gpr20);
+            return {instr.alu_half.type_b, GetRegister(instr.gpr20)};
        default:
            UNREACHABLE();
-            return Immediate(0);
+            return {HalfType::F32, Immediate(0)};
        }
    }();
-    op_b = UnpackHalfFloat(op_b, instr.alu_half.type_b);
+    op_b = UnpackHalfFloat(op_b, type_b);
-    op_b = GetOperandAbsNegHalf(op_b, instr.alu_half.abs_b, negate_b);
+    // redeclaration to avoid a bug in clang with reusing local bindings in lambdas
+    Node op_b_alt = GetOperandAbsNegHalf(op_b, instr.alu_half.abs_b, negate_b);
    Node value = [&]() {
        switch (opcode->get().GetId()) {
        case OpCode::Id::HADD2_C:
        case OpCode::Id::HADD2_R:
-            return Operation(OperationCode::HAdd, PRECISE, op_a, op_b);
+            return Operation(OperationCode::HAdd, PRECISE, op_a, op_b_alt);
        case OpCode::Id::HMUL2_C:
        case OpCode::Id::HMUL2_R:
-            return Operation(OperationCode::HMul, PRECISE, op_a, op_b);
+            return Operation(OperationCode::HMul, PRECISE, op_a, op_b_alt);
        default:
            UNIMPLEMENTED_MSG("Unhandled half float instruction: {}", opcode->get().GetName());
            return Immediate(0);
        }
    }();
+    value = GetSaturatedHalfFloat(value, instr.alu_half.saturate);
    value = HalfMerge(GetRegister(instr.gpr0), value, instr.alu_half.merge);
    SetRegister(bb, instr.gpr0, value);
@@ -68,4 +70,4 @@ u32 ShaderIR::DecodeArithmeticHalf(NodeBlock& bb, u32 pc) {
    return pc;
 }
-} // namespace VideoCommon::Shader
-\ No newline at end of file
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index ba15b1115..b5ec9a6f5 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -120,10 +120,11 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
                return Operation(OperationCode::FCeil, PRECISE, value);
            case Tegra::Shader::F2fRoundingOp::Trunc:
                return Operation(OperationCode::FTrunc, PRECISE, value);
+            default:
+                UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
+                                  static_cast<u32>(instr.conversion.f2f.rounding.Value()));
+                return Immediate(0);
            }
-            UNIMPLEMENTED_MSG("Unimplemented F2F rounding mode {}",
-                              static_cast<u32>(instr.conversion.f2f.rounding.Value()));
-            return Immediate(0);
        }();
        value = GetSaturatedFloat(value, instr.alu.saturate_d);
diff --git a/src/video_core/shader/decode/hfma2.cpp b/src/video_core/shader/decode/hfma2.cpp
index 5c1becce5..a425f9eb7 100644
--- a/src/video_core/shader/decode/hfma2.cpp
+++ b/src/video_core/shader/decode/hfma2.cpp
@@ -34,15 +34,14 @@ u32 ShaderIR::DecodeHfma2(NodeBlock& bb, u32 pc) {
        case OpCode::Id::HFMA2_CR:
            neg_b = instr.hfma2.negate_b;
            neg_c = instr.hfma2.negate_c;
-            return {instr.hfma2.saturate, instr.hfma2.type_b,
+            return {instr.hfma2.saturate, HalfType::F32,
                    GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset()),
                    instr.hfma2.type_reg39, GetRegister(instr.gpr39)};
        case OpCode::Id::HFMA2_RC:
            neg_b = instr.hfma2.negate_b;
            neg_c = instr.hfma2.negate_c;
            return {instr.hfma2.saturate, instr.hfma2.type_reg39, GetRegister(instr.gpr39),
-                    instr.hfma2.type_b,
+                    HalfType::F32, GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset())};
-                    GetConstBuffer(instr.cbuf34.index, instr.cbuf34.GetOffset())};
        case OpCode::Id::HFMA2_RR:
            neg_b = instr.hfma2.rr.negate_b;
            neg_c = instr.hfma2.rr.negate_c;
@@ -56,13 +55,13 @@ u32 ShaderIR::DecodeHfma2(NodeBlock& bb, u32 pc) {
            return {false, identity, Immediate(0), identity, Immediate(0)};
        }
    }();
-    UNIMPLEMENTED_IF_MSG(saturate, "HFMA2 saturation is not implemented");
    const Node op_a = UnpackHalfFloat(GetRegister(instr.gpr8), instr.hfma2.type_a);
    op_b = GetOperandAbsNegHalf(UnpackHalfFloat(op_b, type_b), false, neg_b);
    op_c = GetOperandAbsNegHalf(UnpackHalfFloat(op_c, type_c), false, neg_c);
    Node value = Operation(OperationCode::HFma, PRECISE, op_a, op_b, op_c);
+    value = GetSaturatedHalfFloat(value, saturate);
    value = HalfMerge(GetRegister(instr.gpr0), value, instr.hfma2.merge);
    SetRegister(bb, instr.gpr0, value);
@@ -70,4 +69,4 @@ u32 ShaderIR::DecodeHfma2(NodeBlock& bb, u32 pc) {
    return pc;
 }
-} // namespace VideoCommon::Shader
-\ No newline at end of file
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index fa65ac9a9..8b574d4e5 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -296,7 +296,7 @@ const Sampler& ShaderIR::GetBindlessSampler(const Tegra::Shader::Register& reg,
    ASSERT(cbuf_offset_imm != nullptr);
    const auto cbuf_offset = cbuf_offset_imm->GetValue();
    const auto cbuf_index = cbuf->GetIndex();
-    const u64 cbuf_key = (cbuf_index << 32) | cbuf_offset;
+    const auto cbuf_key = (static_cast<u64>(cbuf_index) << 32) | static_cast<u64>(cbuf_offset);
    // If this sampler has already been used, return the existing mapping.
    const auto itr =
@@ -541,7 +541,6 @@ Node4 ShaderIR::GetTld4Code(Instruction instr, TextureType texture_type, bool de
                            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/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp
index db15c0718..04a776398 100644
--- a/src/video_core/shader/decode/xmad.cpp
+++ b/src/video_core/shader/decode/xmad.cpp
@@ -56,9 +56,10 @@ u32 ShaderIR::DecodeXmad(NodeBlock& bb, u32 pc) {
                    instr.xmad.mode,
                    Immediate(static_cast<u32>(instr.xmad.imm20_16)),
                    GetRegister(instr.gpr39)};
+        default:
+            UNIMPLEMENTED_MSG("Unhandled XMAD instruction: {}", opcode->get().GetName());
+            return {false, false, false, Tegra::Shader::XmadMode::None, Immediate(0), Immediate(0)};
        }
-        UNIMPLEMENTED_MSG("Unhandled XMAD instruction: {}", opcode->get().GetName());
-        return {false, false, false, Tegra::Shader::XmadMode::None, Immediate(0), Immediate(0)};
    }();
    op_a = BitfieldExtract(op_a, instr.xmad.high_a ? 16 : 0, 16);
diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp
index 17f2f711c..e4eb0dfd9 100644
--- a/src/video_core/shader/shader_ir.cpp
+++ b/src/video_core/shader/shader_ir.cpp
@@ -439,11 +439,14 @@ Node ShaderIR::BitfieldExtract(Node value, u32 offset, u32 bits) {
        return OperationCode::LogicalUGreaterEqual;
    case OperationCode::INegate:
        UNREACHABLE_MSG("Can't negate an unsigned integer");
+        return {};
    case OperationCode::IAbsolute:
        UNREACHABLE_MSG("Can't apply absolute to an unsigned integer");
+        return {};
+    default:
+        UNREACHABLE_MSG("Unknown signed operation with code={}", static_cast<u32>(operation_code));
+        return {};
    }
-    UNREACHABLE_MSG("Unknown signed operation with code={}", static_cast<u32>(operation_code));
-    return {};
 }
-} // namespace VideoCommon::Shader
-\ No newline at end of file
+} // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h
index 81278fb33..65f1e1de9 100644
--- a/src/video_core/shader/shader_ir.h
+++ b/src/video_core/shader/shader_ir.h
@@ -251,8 +251,9 @@ public:
    }
    bool operator<(const Sampler& rhs) const {
-        return std::tie(offset, index, type, is_array, is_shadow) <
+        return std::tie(index, offset, type, is_array, is_shadow, is_bindless) <
-               std::tie(rhs.offset, rhs.index, rhs.type, rhs.is_array, rhs.is_shadow);
+               std::tie(rhs.index, rhs.offset, rhs.type, rhs.is_array, rhs.is_shadow,
+                        rhs.is_bindless);
    }
 private:
diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp
index 3b022a456..6384fa8d2 100644
--- a/src/video_core/surface.cpp
+++ b/src/video_core/surface.cpp
@@ -178,39 +178,44 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format,
            return PixelFormat::ABGR8S;
        case Tegra::Texture::ComponentType::UINT:
            return PixelFormat::ABGR8UI;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::B5G6R5:
        switch (component_type) {
        case Tegra::Texture::ComponentType::UNORM:
            return PixelFormat::B5G6R5U;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::A2B10G10R10:
        switch (component_type) {
        case Tegra::Texture::ComponentType::UNORM:
            return PixelFormat::A2B10G10R10U;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::A1B5G5R5:
        switch (component_type) {
        case Tegra::Texture::ComponentType::UNORM:
            return PixelFormat::A1B5G5R5U;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::R8:
        switch (component_type) {
        case Tegra::Texture::ComponentType::UNORM:
            return PixelFormat::R8U;
        case Tegra::Texture::ComponentType::UINT:
            return PixelFormat::R8UI;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::G8R8:
        // TextureFormat::G8R8 is actually ordered red then green, as such we can use
        // PixelFormat::RG8U and PixelFormat::RG8S. This was tested with The Legend of Zelda: Breath
@@ -220,50 +225,55 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format,
            return PixelFormat::RG8U;
        case Tegra::Texture::ComponentType::SNORM:
            return PixelFormat::RG8S;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::R16_G16_B16_A16:
        switch (component_type) {
        case Tegra::Texture::ComponentType::UNORM:
            return PixelFormat::RGBA16U;
        case Tegra::Texture::ComponentType::FLOAT:
            return PixelFormat::RGBA16F;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::BF10GF11RF11:
        switch (component_type) {
        case Tegra::Texture::ComponentType::FLOAT:
            return PixelFormat::R11FG11FB10F;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::R32_G32_B32_A32:
        switch (component_type) {
        case Tegra::Texture::ComponentType::FLOAT:
            return PixelFormat::RGBA32F;
        case Tegra::Texture::ComponentType::UINT:
            return PixelFormat::RGBA32UI;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::R32_G32:
        switch (component_type) {
        case Tegra::Texture::ComponentType::FLOAT:
            return PixelFormat::RG32F;
        case Tegra::Texture::ComponentType::UINT:
            return PixelFormat::RG32UI;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::R32_G32_B32:
        switch (component_type) {
        case Tegra::Texture::ComponentType::FLOAT:
            return PixelFormat::RGB32F;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::R16:
        switch (component_type) {
        case Tegra::Texture::ComponentType::FLOAT:
@@ -276,18 +286,20 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format,
            return PixelFormat::R16UI;
        case Tegra::Texture::ComponentType::SINT:
            return PixelFormat::R16I;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::R32:
        switch (component_type) {
        case Tegra::Texture::ComponentType::FLOAT:
            return PixelFormat::R32F;
        case Tegra::Texture::ComponentType::UINT:
            return PixelFormat::R32UI;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::ZF32:
        return PixelFormat::Z32F;
    case Tegra::Texture::TextureFormat::Z16:
@@ -310,9 +322,10 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format,
            return PixelFormat::DXN2UNORM;
        case Tegra::Texture::ComponentType::SNORM:
            return PixelFormat::DXN2SNORM;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    case Tegra::Texture::TextureFormat::BC7U:
        return is_srgb ? PixelFormat::BC7U_SRGB : PixelFormat::BC7U;
    case Tegra::Texture::TextureFormat::BC6H_UF16:
@@ -343,15 +356,17 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format,
            return PixelFormat::RG16UI;
        case Tegra::Texture::ComponentType::SINT:
            return PixelFormat::RG16I;
+        default:
+            break;
        }
-        LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
+        break;
-        UNREACHABLE();
    default:
-        LOG_CRITICAL(HW_GPU, "Unimplemented format={}, component_type={}", static_cast<u32>(format),
+        break;
-                     static_cast<u32>(component_type));
-        UNREACHABLE();
-        return PixelFormat::ABGR8U;
    }
+    LOG_CRITICAL(HW_GPU, "Unimplemented format={}, component_type={}", static_cast<u32>(format),
+                 static_cast<u32>(component_type));
+    UNREACHABLE();
+    return PixelFormat::ABGR8U;
 }
 ComponentType ComponentTypeFromTexture(Tegra::Texture::ComponentType type) {
@@ -513,8 +528,9 @@ bool IsFormatBCn(PixelFormat format) {
    case PixelFormat::DXT45_SRGB:
    case PixelFormat::BC7U_SRGB:
        return true;
+    default:
+        return false;
    }
-    return false;
 }
 } // namespace VideoCore::Surface
diff --git a/src/video_core/textures/astc.cpp b/src/video_core/textures/astc.cpp
index b508d64e9..eafb6b73a 100644
--- a/src/video_core/textures/astc.cpp
+++ b/src/video_core/textures/astc.cpp
@@ -1616,6 +1616,7 @@ 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 +1631,7 @@ 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 +1639,7 @@ 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/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;