83 files changed, 2167 insertions, 784 deletions
diff --git a/README.md b/README.md
index bdee2e872..e8e7946de 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,6 @@
 yuzu emulator
 =============
-[![Travis CI Build Status](https://travis-ci.org/yuzu-emu/yuzu.svg?branch=master)](https://travis-ci.org/yuzu-emu/yuzu)
+[![Travis CI Build Status](https://travis-ci.com/yuzu-emu/yuzu.svg?branch=master)](https://travis-ci.com/yuzu-emu/yuzu)
 [![Azure Mainline CI Build Status](https://dev.azure.com/yuzu-emu/yuzu/_apis/build/status/yuzu%20mainline?branchName=master)](https://dev.azure.com/yuzu-emu/yuzu/)
 yuzu is an experimental open-source emulator for the Nintendo Switch from the creators of [Citra](https://citra-emu.org/).
@@ -21,7 +21,7 @@ For development discussion, please join us on [Discord](https://discord.gg/XQV6d
 Most of the development happens on GitHub. It's also where [our central repository](https://github.com/yuzu-emu/yuzu) is hosted.
-If you want to contribute please take a look at the [Contributor's Guide](CONTRIBUTING.md) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should as well contact any of the developers on Discord in order to know about the current state of the emulator.
+If you want to contribute please take a look at the [Contributor's Guide](CONTRIBUTING.md) and [Developer Information](https://github.com/yuzu-emu/yuzu/wiki/Developer-Information). You should also contact any of the developers on Discord in order to know about the current state of the emulator.
 ### Building
diff --git a/src/audio_core/algorithm/interpolate.cpp b/src/audio_core/algorithm/interpolate.cpp
index a58f24169..49ab9d3e1 100644
--- a/src/audio_core/algorithm/interpolate.cpp
+++ b/src/audio_core/algorithm/interpolate.cpp
@@ -8,13 +8,14 @@
 #include <climits>
 #include <cmath>
 #include <vector>
 #include "audio_core/algorithm/interpolate.h"
 #include "common/common_types.h"
 #include "common/logging/log.h"
 namespace AudioCore {
-constexpr std::array<s16, 512> curve_lut0 = {
+constexpr std::array<s16, 512> curve_lut0{
    6600,  19426, 6722,  3,     6479,  19424, 6845,  9,     6359,  19419, 6968,  15,    6239,
    19412, 7093,  22,    6121,  19403, 7219,  28,    6004,  19391, 7345,  34,    5888,  19377,
    7472,  41,    5773,  19361, 7600,  48,    5659,  19342, 7728,  55,    5546,  19321, 7857,
@@ -56,7 +57,7 @@ constexpr std::array<s16, 512> curve_lut0 = {
    19403, 6121,  22,    7093,  19412, 6239,  15,    6968,  19419, 6359,  9,     6845,  19424,
    6479,  3,     6722,  19426, 6600};
-constexpr std::array<s16, 512> curve_lut1 = {
+constexpr std::array<s16, 512> curve_lut1{
    -68,   32639, 69,    -5,    -200,  32630, 212,   -15,   -328,  32613, 359,   -26,   -450,
    32586, 512,   -36,   -568,  32551, 669,   -47,   -680,  32507, 832,   -58,   -788,  32454,
    1000,  -69,   -891,  32393, 1174,  -80,   -990,  32323, 1352,  -92,   -1084, 32244, 1536,
@@ -98,7 +99,7 @@ constexpr std::array<s16, 512> curve_lut1 = {
    32551, -568,  -36,   512,   32586, -450,  -26,   359,   32613, -328,  -15,   212,   32630,
    -200,  -5,    69,    32639, -68};
-constexpr std::array<s16, 512> curve_lut2 = {
+constexpr std::array<s16, 512> curve_lut2{
    3195,  26287, 3329,  -32,   3064,  26281, 3467,  -34,   2936,  26270, 3608,  -38,   2811,
    26253, 3751,  -42,   2688,  26230, 3897,  -46,   2568,  26202, 4046,  -50,   2451,  26169,
    4199,  -54,   2338,  26130, 4354,  -58,   2227,  26085, 4512,  -63,   2120,  26035, 4673,
@@ -146,10 +147,10 @@ std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input,
    if (ratio <= 0) {
        LOG_CRITICAL(Audio, "Nonsensical interpolation ratio {}", ratio);
-        ratio = 1.0;
+        return input;
    }
-    const int step = static_cast<int>(ratio * 0x8000);
+    const s32 step{static_cast<s32>(ratio * 0x8000)};
    const std::array<s16, 512>& lut = [step] {
        if (step > 0xaaaa) {
            return curve_lut0;
@@ -160,28 +161,37 @@ std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input,
        return curve_lut2;
    }();
-    std::vector<s16> output(static_cast<std::size_t>(input.size() / ratio));
+    const std::size_t num_frames{input.size() / 2};
-    int in_offset = 0;
-    for (std::size_t out_offset = 0; out_offset < output.size(); out_offset += 2) {
+    std::vector<s16> output;
-        const int lut_index = (state.fraction >> 8) * 4;
+    output.reserve(static_cast<std::size_t>(input.size() / ratio + InterpolationState::taps));
-        const int l = input[(in_offset + 0) * 2 + 0] * lut[lut_index + 0] +
+    for (std::size_t frame{}; frame < num_frames; ++frame) {
-                      input[(in_offset + 1) * 2 + 0] * lut[lut_index + 1] +
+        const std::size_t lut_index{(state.fraction >> 8) * InterpolationState::taps};
-                      input[(in_offset + 2) * 2 + 0] * lut[lut_index + 2] +
-                      input[(in_offset + 3) * 2 + 0] * lut[lut_index + 3];
-        const int r = input[(in_offset + 0) * 2 + 1] * lut[lut_index + 0] +
+        std::rotate(state.history.begin(), state.history.end() - 1, state.history.end());
-                      input[(in_offset + 1) * 2 + 1] * lut[lut_index + 1] +
+        state.history[0][0] = input[frame * 2 + 0];
-                      input[(in_offset + 2) * 2 + 1] * lut[lut_index + 2] +
+        state.history[0][1] = input[frame * 2 + 1];
-                      input[(in_offset + 3) * 2 + 1] * lut[lut_index + 3];
-        const int new_offset = state.fraction + step;
+        while (state.position <= 1.0) {
+            const s32 left{state.history[0][0] * lut[lut_index + 0] +
+                           state.history[1][0] * lut[lut_index + 1] +
+                           state.history[2][0] * lut[lut_index + 2] +
+                           state.history[3][0] * lut[lut_index + 3]};
+            const s32 right{state.history[0][1] * lut[lut_index + 0] +
+                            state.history[1][1] * lut[lut_index + 1] +
+                            state.history[2][1] * lut[lut_index + 2] +
+                            state.history[3][1] * lut[lut_index + 3]};
+            const s32 new_offset{state.fraction + step};
-        in_offset += new_offset >> 15;
+            state.fraction = new_offset & 0x7fff;
-        state.fraction = new_offset & 0x7fff;
-        output[out_offset + 0] = static_cast<s16>(std::clamp(l >> 15, SHRT_MIN, SHRT_MAX));
+            output.emplace_back(static_cast<s16>(std::clamp(left >> 15, SHRT_MIN, SHRT_MAX)));
-        output[out_offset + 1] = static_cast<s16>(std::clamp(r >> 15, SHRT_MIN, SHRT_MAX));
+            output.emplace_back(static_cast<s16>(std::clamp(right >> 15, SHRT_MIN, SHRT_MAX)));
+            state.position += ratio;
+        }
+        state.position -= 1.0;
    }
    return output;
diff --git a/src/audio_core/algorithm/interpolate.h b/src/audio_core/algorithm/interpolate.h
index 1b9831a75..ab1a31754 100644
--- a/src/audio_core/algorithm/interpolate.h
+++ b/src/audio_core/algorithm/interpolate.h
@@ -6,12 +6,17 @@
 #include <array>
 #include <vector>
 #include "common/common_types.h"
 namespace AudioCore {
 struct InterpolationState {
-    int fraction = 0;
+    static constexpr std::size_t taps{4};
+    static constexpr std::size_t history_size{taps * 2 - 1};
+    std::array<std::array<s16, 2>, history_size> history{};
+    double position{};
+    s32 fraction{};
 };
 /// Interpolates input signal to produce output signal.
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index 88c06b2ce..b31a0328c 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -131,8 +131,8 @@ add_library(core STATIC
    frontend/framebuffer_layout.cpp
    frontend/framebuffer_layout.h
    frontend/input.h
-    frontend/scope_acquire_window_context.cpp
+    frontend/scope_acquire_context.cpp
-    frontend/scope_acquire_window_context.h
+    frontend/scope_acquire_context.h
    gdbstub/gdbstub.cpp
    gdbstub/gdbstub.h
    hardware_interrupt_manager.cpp
@@ -595,8 +595,12 @@ endif()
 if (ARCHITECTURE_x86_64)
    target_sources(core PRIVATE
-        arm/dynarmic/arm_dynarmic.cpp
+        arm/dynarmic/arm_dynarmic_32.cpp
-        arm/dynarmic/arm_dynarmic.h
+        arm/dynarmic/arm_dynarmic_32.h
+        arm/dynarmic/arm_dynarmic_64.cpp
+        arm/dynarmic/arm_dynarmic_64.h
+        arm/dynarmic/arm_dynarmic_cp15.cpp
+        arm/dynarmic/arm_dynarmic_cp15.h
    )
    target_link_libraries(core PRIVATE dynarmic)
 endif()
diff --git a/src/core/arm/arm_interface.h b/src/core/arm/arm_interface.h
index 47b964eb7..57eae839e 100644
--- a/src/core/arm/arm_interface.h
+++ b/src/core/arm/arm_interface.h
@@ -25,7 +25,20 @@ public:
    explicit ARM_Interface(System& system_) : system{system_} {}
    virtual ~ARM_Interface() = default;
-    struct ThreadContext {
+    struct ThreadContext32 {
+        std::array<u32, 16> cpu_registers;
+        u32 cpsr;
+        std::array<u8, 4> padding;
+        std::array<u64, 32> fprs;
+        u32 fpscr;
+        u32 fpexc;
+        u32 tpidr;
+    };
+    // Internally within the kernel, it expects the AArch32 version of the
+    // thread context to be 344 bytes in size.
+    static_assert(sizeof(ThreadContext32) == 0x158);
+    struct ThreadContext64 {
        std::array<u64, 31> cpu_registers;
        u64 sp;
        u64 pc;
@@ -38,7 +51,7 @@ public:
    };
    // Internally within the kernel, it expects the AArch64 version of the
    // thread context to be 800 bytes in size.
-    static_assert(sizeof(ThreadContext) == 0x320);
+    static_assert(sizeof(ThreadContext64) == 0x320);
    /// Runs the CPU until an event happens
    virtual void Run() = 0;
@@ -130,17 +143,10 @@ public:
     */
    virtual void SetTPIDR_EL0(u64 value) = 0;
-    /**
+    virtual void SaveContext(ThreadContext32& ctx) = 0;
-     * Saves the current CPU context
+    virtual void SaveContext(ThreadContext64& ctx) = 0;
-     * @param ctx Thread context to save
+    virtual void LoadContext(const ThreadContext32& ctx) = 0;
-     */
+    virtual void LoadContext(const ThreadContext64& ctx) = 0;
-    virtual void SaveContext(ThreadContext& ctx) = 0;
-    /**
-     * Loads a CPU context
-     * @param ctx Thread context to load
-     */
-    virtual void LoadContext(const ThreadContext& ctx) = 0;
    /// Clears the exclusive monitor's state.
    virtual void ClearExclusiveState() = 0;
diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.cpp b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
new file mode 100644
index 000000000..187a972ac
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
@@ -0,0 +1,208 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#include <cinttypes>
+#include <memory>
+#include <dynarmic/A32/a32.h>
+#include <dynarmic/A32/config.h>
+#include <dynarmic/A32/context.h>
+#include "common/microprofile.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
+#include "core/arm/dynarmic/arm_dynarmic_cp15.h"
+#include "core/core.h"
+#include "core/core_manager.h"
+#include "core/core_timing.h"
+#include "core/hle/kernel/svc.h"
+#include "core/memory.h"
+namespace Core {
+class DynarmicCallbacks32 : public Dynarmic::A32::UserCallbacks {
+public:
+    explicit DynarmicCallbacks32(ARM_Dynarmic_32& parent) : parent(parent) {}
+    u8 MemoryRead8(u32 vaddr) override {
+        return parent.system.Memory().Read8(vaddr);
+    }
+    u16 MemoryRead16(u32 vaddr) override {
+        return parent.system.Memory().Read16(vaddr);
+    }
+    u32 MemoryRead32(u32 vaddr) override {
+        return parent.system.Memory().Read32(vaddr);
+    }
+    u64 MemoryRead64(u32 vaddr) override {
+        return parent.system.Memory().Read64(vaddr);
+    }
+    void MemoryWrite8(u32 vaddr, u8 value) override {
+        parent.system.Memory().Write8(vaddr, value);
+    }
+    void MemoryWrite16(u32 vaddr, u16 value) override {
+        parent.system.Memory().Write16(vaddr, value);
+    }
+    void MemoryWrite32(u32 vaddr, u32 value) override {
+        parent.system.Memory().Write32(vaddr, value);
+    }
+    void MemoryWrite64(u32 vaddr, u64 value) override {
+        parent.system.Memory().Write64(vaddr, value);
+    }
+    void InterpreterFallback(u32 pc, std::size_t num_instructions) override {
+        UNIMPLEMENTED();
+    }
+    void ExceptionRaised(u32 pc, Dynarmic::A32::Exception exception) override {
+        switch (exception) {
+        case Dynarmic::A32::Exception::UndefinedInstruction:
+        case Dynarmic::A32::Exception::UnpredictableInstruction:
+            break;
+        case Dynarmic::A32::Exception::Breakpoint:
+            break;
+        }
+        LOG_CRITICAL(HW_GPU, "ExceptionRaised(exception = {}, pc = {:08X}, code = {:08X})",
+                     static_cast<std::size_t>(exception), pc, MemoryReadCode(pc));
+        UNIMPLEMENTED();
+    }
+    void CallSVC(u32 swi) override {
+        Kernel::CallSVC(parent.system, swi);
+    }
+    void AddTicks(u64 ticks) override {
+        // Divide the number of ticks by the amount of CPU cores. TODO(Subv): This yields only a
+        // rough approximation of the amount of executed ticks in the system, it may be thrown off
+        // if not all cores are doing a similar amount of work. Instead of doing this, we should
+        // device a way so that timing is consistent across all cores without increasing the ticks 4
+        // times.
+        u64 amortized_ticks = (ticks - num_interpreted_instructions) / Core::NUM_CPU_CORES;
+        // Always execute at least one tick.
+        amortized_ticks = std::max<u64>(amortized_ticks, 1);
+        parent.system.CoreTiming().AddTicks(amortized_ticks);
+        num_interpreted_instructions = 0;
+    }
+    u64 GetTicksRemaining() override {
+        return std::max(parent.system.CoreTiming().GetDowncount(), {});
+    }
+    ARM_Dynarmic_32& parent;
+    std::size_t num_interpreted_instructions{};
+    u64 tpidrro_el0{};
+    u64 tpidr_el0{};
+};
+std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable& page_table,
+                                                             std::size_t address_space_bits) const {
+    Dynarmic::A32::UserConfig config;
+    config.callbacks = cb.get();
+    // TODO(bunnei): Implement page table for 32-bit
+    // config.page_table = &page_table.pointers;
+    config.coprocessors[15] = std::make_shared<DynarmicCP15>((u32*)&CP15_regs[0]);
+    config.define_unpredictable_behaviour = true;
+    return std::make_unique<Dynarmic::A32::Jit>(config);
+}
+MICROPROFILE_DEFINE(ARM_Jit_Dynarmic_32, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
+void ARM_Dynarmic_32::Run() {
+    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic_32);
+    jit->Run();
+}
+void ARM_Dynarmic_32::Step() {
+    cb->InterpreterFallback(jit->Regs()[15], 1);
+}
+ARM_Dynarmic_32::ARM_Dynarmic_32(System& system, ExclusiveMonitor& exclusive_monitor,
+                                 std::size_t core_index)
+    : ARM_Interface{system},
+      cb(std::make_unique<DynarmicCallbacks32>(*this)), core_index{core_index},
+      exclusive_monitor{dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
+ARM_Dynarmic_32::~ARM_Dynarmic_32() = default;
+void ARM_Dynarmic_32::SetPC(u64 pc) {
+    jit->Regs()[15] = static_cast<u32>(pc);
+}
+u64 ARM_Dynarmic_32::GetPC() const {
+    return jit->Regs()[15];
+}
+u64 ARM_Dynarmic_32::GetReg(int index) const {
+    return jit->Regs()[index];
+}
+void ARM_Dynarmic_32::SetReg(int index, u64 value) {
+    jit->Regs()[index] = static_cast<u32>(value);
+}
+u128 ARM_Dynarmic_32::GetVectorReg(int index) const {
+    return {};
+}
+void ARM_Dynarmic_32::SetVectorReg(int index, u128 value) {}
+u32 ARM_Dynarmic_32::GetPSTATE() const {
+    return jit->Cpsr();
+}
+void ARM_Dynarmic_32::SetPSTATE(u32 cpsr) {
+    jit->SetCpsr(cpsr);
+}
+u64 ARM_Dynarmic_32::GetTlsAddress() const {
+    return CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
+}
+void ARM_Dynarmic_32::SetTlsAddress(VAddr address) {
+    CP15_regs[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)] = static_cast<u32>(address);
+}
+u64 ARM_Dynarmic_32::GetTPIDR_EL0() const {
+    return cb->tpidr_el0;
+}
+void ARM_Dynarmic_32::SetTPIDR_EL0(u64 value) {
+    cb->tpidr_el0 = value;
+}
+void ARM_Dynarmic_32::SaveContext(ThreadContext32& ctx) {
+    Dynarmic::A32::Context context;
+    jit->SaveContext(context);
+    ctx.cpu_registers = context.Regs();
+    ctx.cpsr = context.Cpsr();
+}
+void ARM_Dynarmic_32::LoadContext(const ThreadContext32& ctx) {
+    Dynarmic::A32::Context context;
+    context.Regs() = ctx.cpu_registers;
+    context.SetCpsr(ctx.cpsr);
+    jit->LoadContext(context);
+}
+void ARM_Dynarmic_32::PrepareReschedule() {
+    jit->HaltExecution();
+}
+void ARM_Dynarmic_32::ClearInstructionCache() {
+    jit->ClearCache();
+}
+void ARM_Dynarmic_32::ClearExclusiveState() {}
+void ARM_Dynarmic_32::PageTableChanged(Common::PageTable& page_table,
+                                       std::size_t new_address_space_size_in_bits) {
+    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
+    auto iter = jit_cache.find(key);
+    if (iter != jit_cache.end()) {
+        jit = iter->second;
+        return;
+    }
+    jit = MakeJit(page_table, new_address_space_size_in_bits);
+    jit_cache.emplace(key, jit);
+}
+} // namespace Core
diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.h b/src/core/arm/dynarmic/arm_dynarmic_32.h
new file mode 100644
index 000000000..143e46e4d
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_32.h
@@ -0,0 +1,77 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#pragma once
+#include <memory>
+#include <unordered_map>
+#include <dynarmic/A32/a32.h>
+#include <dynarmic/A64/a64.h>
+#include <dynarmic/A64/exclusive_monitor.h>
+#include "common/common_types.h"
+#include "common/hash.h"
+#include "core/arm/arm_interface.h"
+#include "core/arm/exclusive_monitor.h"
+namespace Memory {
+class Memory;
+}
+namespace Core {
+class DynarmicCallbacks32;
+class DynarmicExclusiveMonitor;
+class System;
+class ARM_Dynarmic_32 final : public ARM_Interface {
+public:
+    ARM_Dynarmic_32(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ~ARM_Dynarmic_32() override;
+    void SetPC(u64 pc) override;
+    u64 GetPC() const override;
+    u64 GetReg(int index) const override;
+    void SetReg(int index, u64 value) override;
+    u128 GetVectorReg(int index) const override;
+    void SetVectorReg(int index, u128 value) override;
+    u32 GetPSTATE() const override;
+    void SetPSTATE(u32 pstate) override;
+    void Run() override;
+    void Step() override;
+    VAddr GetTlsAddress() const override;
+    void SetTlsAddress(VAddr address) override;
+    void SetTPIDR_EL0(u64 value) override;
+    u64 GetTPIDR_EL0() const override;
+    void SaveContext(ThreadContext32& ctx) override;
+    void SaveContext(ThreadContext64& ctx) override {}
+    void LoadContext(const ThreadContext32& ctx) override;
+    void LoadContext(const ThreadContext64& ctx) override {}
+    void PrepareReschedule() override;
+    void ClearExclusiveState() override;
+    void ClearInstructionCache() override;
+    void PageTableChanged(Common::PageTable& new_page_table,
+                          std::size_t new_address_space_size_in_bits) override;
+private:
+    std::shared_ptr<Dynarmic::A32::Jit> MakeJit(Common::PageTable& page_table,
+                                                std::size_t address_space_bits) const;
+    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
+    using JitCacheType =
+        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A32::Jit>, Common::PairHash>;
+    friend class DynarmicCallbacks32;
+    std::unique_ptr<DynarmicCallbacks32> cb;
+    JitCacheType jit_cache;
+    std::shared_ptr<Dynarmic::A32::Jit> jit;
+    std::size_t core_index;
+    DynarmicExclusiveMonitor& exclusive_monitor;
+    std::array<u32, 84> CP15_regs{};
+};
+} // namespace Core
diff --git a/src/core/arm/dynarmic/arm_dynarmic.cpp b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
index 29eaf74e5..a53a58ba0 100644
--- a/src/core/arm/dynarmic/arm_dynarmic.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
@@ -8,7 +8,7 @@
 #include <dynarmic/A64/config.h>
 #include "common/logging/log.h"
 #include "common/microprofile.h"
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #include "core/core.h"
 #include "core/core_manager.h"
 #include "core/core_timing.h"
@@ -25,9 +25,9 @@ namespace Core {
 using Vector = Dynarmic::A64::Vector;
-class ARM_Dynarmic_Callbacks : public Dynarmic::A64::UserCallbacks {
+class DynarmicCallbacks64 : public Dynarmic::A64::UserCallbacks {
 public:
-    explicit ARM_Dynarmic_Callbacks(ARM_Dynarmic& parent) : parent(parent) {}
+    explicit DynarmicCallbacks64(ARM_Dynarmic_64& parent) : parent(parent) {}
    u8 MemoryRead8(u64 vaddr) override {
        return parent.system.Memory().Read8(vaddr);
@@ -68,7 +68,7 @@ public:
        LOG_INFO(Core_ARM, "Unicorn fallback @ 0x{:X} for {} instructions (instr = {:08X})", pc,
                 num_instructions, MemoryReadCode(pc));
-        ARM_Interface::ThreadContext ctx;
+        ARM_Interface::ThreadContext64 ctx;
        parent.SaveContext(ctx);
        parent.inner_unicorn.LoadContext(ctx);
        parent.inner_unicorn.ExecuteInstructions(num_instructions);
@@ -90,7 +90,7 @@ public:
                parent.jit->HaltExecution();
                parent.SetPC(pc);
                Kernel::Thread* const thread = parent.system.CurrentScheduler().GetCurrentThread();
-                parent.SaveContext(thread->GetContext());
+                parent.SaveContext(thread->GetContext64());
                GDBStub::Break();
                GDBStub::SendTrap(thread, 5);
                return;
@@ -126,14 +126,14 @@ public:
        return Timing::CpuCyclesToClockCycles(parent.system.CoreTiming().GetTicks());
    }
-    ARM_Dynarmic& parent;
+    ARM_Dynarmic_64& parent;
    std::size_t num_interpreted_instructions = 0;
    u64 tpidrro_el0 = 0;
    u64 tpidr_el0 = 0;
 };
-std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit(Common::PageTable& page_table,
+std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable& page_table,
-                                                          std::size_t address_space_bits) const {
+                                                             std::size_t address_space_bits) const {
    Dynarmic::A64::UserConfig config;
    // Callbacks
@@ -159,79 +159,79 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit(Common::PageTable& pag
    // Unpredictable instructions
    config.define_unpredictable_behaviour = true;
-    return std::make_unique<Dynarmic::A64::Jit>(config);
+    return std::make_shared<Dynarmic::A64::Jit>(config);
 }
-MICROPROFILE_DEFINE(ARM_Jit_Dynarmic, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
+MICROPROFILE_DEFINE(ARM_Jit_Dynarmic_64, "ARM JIT", "Dynarmic", MP_RGB(255, 64, 64));
-void ARM_Dynarmic::Run() {
+void ARM_Dynarmic_64::Run() {
-    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic);
+    MICROPROFILE_SCOPE(ARM_Jit_Dynarmic_64);
    jit->Run();
 }
-void ARM_Dynarmic::Step() {
+void ARM_Dynarmic_64::Step() {
    cb->InterpreterFallback(jit->GetPC(), 1);
 }
-ARM_Dynarmic::ARM_Dynarmic(System& system, ExclusiveMonitor& exclusive_monitor,
+ARM_Dynarmic_64::ARM_Dynarmic_64(System& system, ExclusiveMonitor& exclusive_monitor,
-                           std::size_t core_index)
+                                 std::size_t core_index)
    : ARM_Interface{system},
-      cb(std::make_unique<ARM_Dynarmic_Callbacks>(*this)), inner_unicorn{system},
+      cb(std::make_unique<DynarmicCallbacks64>(*this)), inner_unicorn{system},
      core_index{core_index}, exclusive_monitor{
                                  dynamic_cast<DynarmicExclusiveMonitor&>(exclusive_monitor)} {}
-ARM_Dynarmic::~ARM_Dynarmic() = default;
+ARM_Dynarmic_64::~ARM_Dynarmic_64() = default;
-void ARM_Dynarmic::SetPC(u64 pc) {
+void ARM_Dynarmic_64::SetPC(u64 pc) {
    jit->SetPC(pc);
 }
-u64 ARM_Dynarmic::GetPC() const {
+u64 ARM_Dynarmic_64::GetPC() const {
    return jit->GetPC();
 }
-u64 ARM_Dynarmic::GetReg(int index) const {
+u64 ARM_Dynarmic_64::GetReg(int index) const {
    return jit->GetRegister(index);
 }
-void ARM_Dynarmic::SetReg(int index, u64 value) {
+void ARM_Dynarmic_64::SetReg(int index, u64 value) {
    jit->SetRegister(index, value);
 }
-u128 ARM_Dynarmic::GetVectorReg(int index) const {
+u128 ARM_Dynarmic_64::GetVectorReg(int index) const {
    return jit->GetVector(index);
 }
-void ARM_Dynarmic::SetVectorReg(int index, u128 value) {
+void ARM_Dynarmic_64::SetVectorReg(int index, u128 value) {
    jit->SetVector(index, value);
 }
-u32 ARM_Dynarmic::GetPSTATE() const {
+u32 ARM_Dynarmic_64::GetPSTATE() const {
    return jit->GetPstate();
 }
-void ARM_Dynarmic::SetPSTATE(u32 pstate) {
+void ARM_Dynarmic_64::SetPSTATE(u32 pstate) {
    jit->SetPstate(pstate);
 }
-u64 ARM_Dynarmic::GetTlsAddress() const {
+u64 ARM_Dynarmic_64::GetTlsAddress() const {
    return cb->tpidrro_el0;
 }
-void ARM_Dynarmic::SetTlsAddress(VAddr address) {
+void ARM_Dynarmic_64::SetTlsAddress(VAddr address) {
    cb->tpidrro_el0 = address;
 }
-u64 ARM_Dynarmic::GetTPIDR_EL0() const {
+u64 ARM_Dynarmic_64::GetTPIDR_EL0() const {
    return cb->tpidr_el0;
 }
-void ARM_Dynarmic::SetTPIDR_EL0(u64 value) {
+void ARM_Dynarmic_64::SetTPIDR_EL0(u64 value) {
    cb->tpidr_el0 = value;
 }
-void ARM_Dynarmic::SaveContext(ThreadContext& ctx) {
+void ARM_Dynarmic_64::SaveContext(ThreadContext64& ctx) {
    ctx.cpu_registers = jit->GetRegisters();
    ctx.sp = jit->GetSP();
    ctx.pc = jit->GetPC();
@@ -242,7 +242,7 @@ void ARM_Dynarmic::SaveContext(ThreadContext& ctx) {
    ctx.tpidr = cb->tpidr_el0;
 }
-void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
+void ARM_Dynarmic_64::LoadContext(const ThreadContext64& ctx) {
    jit->SetRegisters(ctx.cpu_registers);
    jit->SetSP(ctx.sp);
    jit->SetPC(ctx.pc);
@@ -253,25 +253,32 @@ void ARM_Dynarmic::LoadContext(const ThreadContext& ctx) {
    SetTPIDR_EL0(ctx.tpidr);
 }
-void ARM_Dynarmic::PrepareReschedule() {
+void ARM_Dynarmic_64::PrepareReschedule() {
    jit->HaltExecution();
 }
-void ARM_Dynarmic::ClearInstructionCache() {
+void ARM_Dynarmic_64::ClearInstructionCache() {
    jit->ClearCache();
 }
-void ARM_Dynarmic::ClearExclusiveState() {
+void ARM_Dynarmic_64::ClearExclusiveState() {
    jit->ClearExclusiveState();
 }
-void ARM_Dynarmic::PageTableChanged(Common::PageTable& page_table,
+void ARM_Dynarmic_64::PageTableChanged(Common::PageTable& page_table,
-                                    std::size_t new_address_space_size_in_bits) {
+                                       std::size_t new_address_space_size_in_bits) {
+    auto key = std::make_pair(&page_table, new_address_space_size_in_bits);
+    auto iter = jit_cache.find(key);
+    if (iter != jit_cache.end()) {
+        jit = iter->second;
+        return;
+    }
    jit = MakeJit(page_table, new_address_space_size_in_bits);
+    jit_cache.emplace(key, jit);
 }
-DynarmicExclusiveMonitor::DynarmicExclusiveMonitor(Memory::Memory& memory_, std::size_t core_count)
+DynarmicExclusiveMonitor::DynarmicExclusiveMonitor(Memory::Memory& memory, std::size_t core_count)
-    : monitor(core_count), memory{memory_} {}
+    : monitor(core_count), memory{memory} {}
 DynarmicExclusiveMonitor::~DynarmicExclusiveMonitor() = default;
diff --git a/src/core/arm/dynarmic/arm_dynarmic.h b/src/core/arm/dynarmic/arm_dynarmic_64.h
index 9cd475cfb..e71240a96 100644
--- a/src/core/arm/dynarmic/arm_dynarmic.h
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.h
@@ -5,9 +5,12 @@
 #pragma once
 #include <memory>
+#include <unordered_map>
 #include <dynarmic/A64/a64.h>
 #include <dynarmic/A64/exclusive_monitor.h>
 #include "common/common_types.h"
+#include "common/hash.h"
 #include "core/arm/arm_interface.h"
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
@@ -18,14 +21,14 @@ class Memory;
 namespace Core {
-class ARM_Dynarmic_Callbacks;
+class DynarmicCallbacks64;
 class DynarmicExclusiveMonitor;
 class System;
-class ARM_Dynarmic final : public ARM_Interface {
+class ARM_Dynarmic_64 final : public ARM_Interface {
 public:
-    ARM_Dynarmic(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
+    ARM_Dynarmic_64(System& system, ExclusiveMonitor& exclusive_monitor, std::size_t core_index);
-    ~ARM_Dynarmic() override;
+    ~ARM_Dynarmic_64() override;
    void SetPC(u64 pc) override;
    u64 GetPC() const override;
@@ -42,8 +45,10 @@ public:
    void SetTPIDR_EL0(u64 value) override;
    u64 GetTPIDR_EL0() const override;
-    void SaveContext(ThreadContext& ctx) override;
+    void SaveContext(ThreadContext32& ctx) override {}
-    void LoadContext(const ThreadContext& ctx) override;
+    void SaveContext(ThreadContext64& ctx) override;
+    void LoadContext(const ThreadContext32& ctx) override {}
+    void LoadContext(const ThreadContext64& ctx) override;
    void PrepareReschedule() override;
    void ClearExclusiveState() override;
@@ -53,12 +58,17 @@ public:
                          std::size_t new_address_space_size_in_bits) override;
 private:
-    std::unique_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable& page_table,
+    std::shared_ptr<Dynarmic::A64::Jit> MakeJit(Common::PageTable& page_table,
                                                std::size_t address_space_bits) const;
-    friend class ARM_Dynarmic_Callbacks;
+    using JitCacheKey = std::pair<Common::PageTable*, std::size_t>;
-    std::unique_ptr<ARM_Dynarmic_Callbacks> cb;
+    using JitCacheType =
-    std::unique_ptr<Dynarmic::A64::Jit> jit;
+        std::unordered_map<JitCacheKey, std::shared_ptr<Dynarmic::A64::Jit>, Common::PairHash>;
+    friend class DynarmicCallbacks64;
+    std::unique_ptr<DynarmicCallbacks64> cb;
+    JitCacheType jit_cache;
+    std::shared_ptr<Dynarmic::A64::Jit> jit;
    ARM_Unicorn inner_unicorn;
    std::size_t core_index;
@@ -67,7 +77,7 @@ private:
 class DynarmicExclusiveMonitor final : public ExclusiveMonitor {
 public:
-    explicit DynarmicExclusiveMonitor(Memory::Memory& memory_, std::size_t core_count);
+    explicit DynarmicExclusiveMonitor(Memory::Memory& memory, std::size_t core_count);
    ~DynarmicExclusiveMonitor() override;
    void SetExclusive(std::size_t core_index, VAddr addr) override;
@@ -80,7 +90,7 @@ public:
    bool ExclusiveWrite128(std::size_t core_index, VAddr vaddr, u128 value) override;
 private:
-    friend class ARM_Dynarmic;
+    friend class ARM_Dynarmic_64;
    Dynarmic::A64::ExclusiveMonitor monitor;
    Memory::Memory& memory;
 };
diff --git a/src/core/arm/dynarmic/arm_dynarmic_cp15.cpp b/src/core/arm/dynarmic/arm_dynarmic_cp15.cpp
new file mode 100644
index 000000000..3fdcdebde
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_cp15.cpp
@@ -0,0 +1,80 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#include "core/arm/dynarmic/arm_dynarmic_cp15.h"
+using Callback = Dynarmic::A32::Coprocessor::Callback;
+using CallbackOrAccessOneWord = Dynarmic::A32::Coprocessor::CallbackOrAccessOneWord;
+using CallbackOrAccessTwoWords = Dynarmic::A32::Coprocessor::CallbackOrAccessTwoWords;
+std::optional<Callback> DynarmicCP15::CompileInternalOperation(bool two, unsigned opc1,
+                                                               CoprocReg CRd, CoprocReg CRn,
+                                                               CoprocReg CRm, unsigned opc2) {
+    return {};
+}
+CallbackOrAccessOneWord DynarmicCP15::CompileSendOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                                         CoprocReg CRm, unsigned opc2) {
+    // TODO(merry): Privileged CP15 registers
+    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C5 && opc2 == 4) {
+        // This is a dummy write, we ignore the value written here.
+        return &CP15[static_cast<std::size_t>(CP15Register::CP15_FLUSH_PREFETCH_BUFFER)];
+    }
+    if (!two && CRn == CoprocReg::C7 && opc1 == 0 && CRm == CoprocReg::C10) {
+        switch (opc2) {
+        case 4:
+            // This is a dummy write, we ignore the value written here.
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_SYNC_BARRIER)];
+        case 5:
+            // This is a dummy write, we ignore the value written here.
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_DATA_MEMORY_BARRIER)];
+        default:
+            return {};
+        }
+    }
+    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0 && opc2 == 2) {
+        return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+    }
+    return {};
+}
+CallbackOrAccessTwoWords DynarmicCP15::CompileSendTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    return {};
+}
+CallbackOrAccessOneWord DynarmicCP15::CompileGetOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                                        CoprocReg CRm, unsigned opc2) {
+    // TODO(merry): Privileged CP15 registers
+    if (!two && CRn == CoprocReg::C13 && opc1 == 0 && CRm == CoprocReg::C0) {
+        switch (opc2) {
+        case 2:
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_UPRW)];
+        case 3:
+            return &CP15[static_cast<std::size_t>(CP15Register::CP15_THREAD_URO)];
+        default:
+            return {};
+        }
+    }
+    return {};
+}
+CallbackOrAccessTwoWords DynarmicCP15::CompileGetTwoWords(bool two, unsigned opc, CoprocReg CRm) {
+    return {};
+}
+std::optional<Callback> DynarmicCP15::CompileLoadWords(bool two, bool long_transfer, CoprocReg CRd,
+                                                       std::optional<u8> option) {
+    return {};
+}
+std::optional<Callback> DynarmicCP15::CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
+                                                        std::optional<u8> option) {
+    return {};
+}
diff --git a/src/core/arm/dynarmic/arm_dynarmic_cp15.h b/src/core/arm/dynarmic/arm_dynarmic_cp15.h
new file mode 100644
index 000000000..07bcde5f9
--- /dev/null
+++ b/src/core/arm/dynarmic/arm_dynarmic_cp15.h
@@ -0,0 +1,152 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#pragma once
+#include <memory>
+#include <optional>
+#include <dynarmic/A32/coprocessor.h>
+#include "common/common_types.h"
+enum class CP15Register {
+    // c0 - Information registers
+    CP15_MAIN_ID,
+    CP15_CACHE_TYPE,
+    CP15_TCM_STATUS,
+    CP15_TLB_TYPE,
+    CP15_CPU_ID,
+    CP15_PROCESSOR_FEATURE_0,
+    CP15_PROCESSOR_FEATURE_1,
+    CP15_DEBUG_FEATURE_0,
+    CP15_AUXILIARY_FEATURE_0,
+    CP15_MEMORY_MODEL_FEATURE_0,
+    CP15_MEMORY_MODEL_FEATURE_1,
+    CP15_MEMORY_MODEL_FEATURE_2,
+    CP15_MEMORY_MODEL_FEATURE_3,
+    CP15_ISA_FEATURE_0,
+    CP15_ISA_FEATURE_1,
+    CP15_ISA_FEATURE_2,
+    CP15_ISA_FEATURE_3,
+    CP15_ISA_FEATURE_4,
+    // c1 - Control registers
+    CP15_CONTROL,
+    CP15_AUXILIARY_CONTROL,
+    CP15_COPROCESSOR_ACCESS_CONTROL,
+    // c2 - Translation table registers
+    CP15_TRANSLATION_BASE_TABLE_0,
+    CP15_TRANSLATION_BASE_TABLE_1,
+    CP15_TRANSLATION_BASE_CONTROL,
+    CP15_DOMAIN_ACCESS_CONTROL,
+    CP15_RESERVED,
+    // c5 - Fault status registers
+    CP15_FAULT_STATUS,
+    CP15_INSTR_FAULT_STATUS,
+    CP15_COMBINED_DATA_FSR = CP15_FAULT_STATUS,
+    CP15_INST_FSR,
+    // c6 - Fault Address registers
+    CP15_FAULT_ADDRESS,
+    CP15_COMBINED_DATA_FAR = CP15_FAULT_ADDRESS,
+    CP15_WFAR,
+    CP15_IFAR,
+    // c7 - Cache operation registers
+    CP15_WAIT_FOR_INTERRUPT,
+    CP15_PHYS_ADDRESS,
+    CP15_INVALIDATE_INSTR_CACHE,
+    CP15_INVALIDATE_INSTR_CACHE_USING_MVA,
+    CP15_INVALIDATE_INSTR_CACHE_USING_INDEX,
+    CP15_FLUSH_PREFETCH_BUFFER,
+    CP15_FLUSH_BRANCH_TARGET_CACHE,
+    CP15_FLUSH_BRANCH_TARGET_CACHE_ENTRY,
+    CP15_INVALIDATE_DATA_CACHE,
+    CP15_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
+    CP15_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
+    CP15_INVALIDATE_DATA_AND_INSTR_CACHE,
+    CP15_CLEAN_DATA_CACHE,
+    CP15_CLEAN_DATA_CACHE_LINE_USING_MVA,
+    CP15_CLEAN_DATA_CACHE_LINE_USING_INDEX,
+    CP15_DATA_SYNC_BARRIER,
+    CP15_DATA_MEMORY_BARRIER,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_MVA,
+    CP15_CLEAN_AND_INVALIDATE_DATA_CACHE_LINE_USING_INDEX,
+    // c8 - TLB operations
+    CP15_INVALIDATE_ITLB,
+    CP15_INVALIDATE_ITLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_ITLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_ITLB_ENTRY_ON_MVA,
+    CP15_INVALIDATE_DTLB,
+    CP15_INVALIDATE_DTLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_DTLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_DTLB_ENTRY_ON_MVA,
+    CP15_INVALIDATE_UTLB,
+    CP15_INVALIDATE_UTLB_SINGLE_ENTRY,
+    CP15_INVALIDATE_UTLB_ENTRY_ON_ASID_MATCH,
+    CP15_INVALIDATE_UTLB_ENTRY_ON_MVA,
+    // c9 - Data cache lockdown register
+    CP15_DATA_CACHE_LOCKDOWN,
+    // c10 - TLB/Memory map registers
+    CP15_TLB_LOCKDOWN,
+    CP15_PRIMARY_REGION_REMAP,
+    CP15_NORMAL_REGION_REMAP,
+    // c13 - Thread related registers
+    CP15_PID,
+    CP15_CONTEXT_ID,
+    CP15_THREAD_UPRW, // Thread ID register - User/Privileged Read/Write
+    CP15_THREAD_URO,  // Thread ID register - User Read Only (Privileged R/W)
+    CP15_THREAD_PRW,  // Thread ID register - Privileged R/W only.
+    // c15 - Performance and TLB lockdown registers
+    CP15_PERFORMANCE_MONITOR_CONTROL,
+    CP15_CYCLE_COUNTER,
+    CP15_COUNT_0,
+    CP15_COUNT_1,
+    CP15_READ_MAIN_TLB_LOCKDOWN_ENTRY,
+    CP15_WRITE_MAIN_TLB_LOCKDOWN_ENTRY,
+    CP15_MAIN_TLB_LOCKDOWN_VIRT_ADDRESS,
+    CP15_MAIN_TLB_LOCKDOWN_PHYS_ADDRESS,
+    CP15_MAIN_TLB_LOCKDOWN_ATTRIBUTE,
+    CP15_TLB_DEBUG_CONTROL,
+    // Skyeye defined
+    CP15_TLB_FAULT_ADDR,
+    CP15_TLB_FAULT_STATUS,
+    // Not an actual register.
+    // All registers should be defined above this.
+    CP15_REGISTER_COUNT,
+};
+class DynarmicCP15 final : public Dynarmic::A32::Coprocessor {
+public:
+    using CoprocReg = Dynarmic::A32::CoprocReg;
+    explicit DynarmicCP15(u32* cp15) : CP15(cp15){};
+    std::optional<Callback> CompileInternalOperation(bool two, unsigned opc1, CoprocReg CRd,
+                                                     CoprocReg CRn, CoprocReg CRm,
+                                                     unsigned opc2) override;
+    CallbackOrAccessOneWord CompileSendOneWord(bool two, unsigned opc1, CoprocReg CRn,
+                                               CoprocReg CRm, unsigned opc2) override;
+    CallbackOrAccessTwoWords CompileSendTwoWords(bool two, unsigned opc, CoprocReg CRm) override;
+    CallbackOrAccessOneWord CompileGetOneWord(bool two, unsigned opc1, CoprocReg CRn, CoprocReg CRm,
+                                              unsigned opc2) override;
+    CallbackOrAccessTwoWords CompileGetTwoWords(bool two, unsigned opc, CoprocReg CRm) override;
+    std::optional<Callback> CompileLoadWords(bool two, bool long_transfer, CoprocReg CRd,
+                                             std::optional<u8> option) override;
+    std::optional<Callback> CompileStoreWords(bool two, bool long_transfer, CoprocReg CRd,
+                                              std::optional<u8> option) override;
+private:
+    u32* CP15{};
+};
diff --git a/src/core/arm/exclusive_monitor.cpp b/src/core/arm/exclusive_monitor.cpp
index 94570e520..b32401e0b 100644
--- a/src/core/arm/exclusive_monitor.cpp
+++ b/src/core/arm/exclusive_monitor.cpp
@@ -3,7 +3,7 @@
 // Refer to the license.txt file included.
 #ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/memory.h"
diff --git a/src/core/arm/unicorn/arm_unicorn.cpp b/src/core/arm/unicorn/arm_unicorn.cpp
index f99ad5802..8a9800a96 100644
--- a/src/core/arm/unicorn/arm_unicorn.cpp
+++ b/src/core/arm/unicorn/arm_unicorn.cpp
@@ -53,7 +53,7 @@ static bool UnmappedMemoryHook(uc_engine* uc, uc_mem_type type, u64 addr, int si
                               void* user_data) {
    auto* const system = static_cast<System*>(user_data);
-    ARM_Interface::ThreadContext ctx{};
+    ARM_Interface::ThreadContext64 ctx{};
    system->CurrentArmInterface().SaveContext(ctx);
    ASSERT_MSG(false, "Attempted to read from unmapped memory: 0x{:X}, pc=0x{:X}, lr=0x{:X}", addr,
               ctx.pc, ctx.cpu_registers[30]);
@@ -179,7 +179,7 @@ void ARM_Unicorn::ExecuteInstructions(std::size_t num_instructions) {
        }
        Kernel::Thread* const thread = system.CurrentScheduler().GetCurrentThread();
-        SaveContext(thread->GetContext());
+        SaveContext(thread->GetContext64());
        if (last_bkpt_hit || GDBStub::IsMemoryBreak() || GDBStub::GetCpuStepFlag()) {
            last_bkpt_hit = false;
            GDBStub::Break();
@@ -188,7 +188,7 @@ void ARM_Unicorn::ExecuteInstructions(std::size_t num_instructions) {
    }
 }
-void ARM_Unicorn::SaveContext(ThreadContext& ctx) {
+void ARM_Unicorn::SaveContext(ThreadContext64& ctx) {
    int uregs[32];
    void* tregs[32];
@@ -215,7 +215,7 @@ void ARM_Unicorn::SaveContext(ThreadContext& ctx) {
    CHECKED(uc_reg_read_batch(uc, uregs, tregs, 32));
 }
-void ARM_Unicorn::LoadContext(const ThreadContext& ctx) {
+void ARM_Unicorn::LoadContext(const ThreadContext64& ctx) {
    int uregs[32];
    void* tregs[32];
diff --git a/src/core/arm/unicorn/arm_unicorn.h b/src/core/arm/unicorn/arm_unicorn.h
index 3c5b155f9..f30d13cb6 100644
--- a/src/core/arm/unicorn/arm_unicorn.h
+++ b/src/core/arm/unicorn/arm_unicorn.h
@@ -30,8 +30,6 @@ public:
    void SetTlsAddress(VAddr address) override;
    void SetTPIDR_EL0(u64 value) override;
    u64 GetTPIDR_EL0() const override;
-    void SaveContext(ThreadContext& ctx) override;
-    void LoadContext(const ThreadContext& ctx) override;
    void PrepareReschedule() override;
    void ClearExclusiveState() override;
    void ExecuteInstructions(std::size_t num_instructions);
@@ -41,6 +39,11 @@ public:
    void PageTableChanged(Common::PageTable&, std::size_t) override {}
    void RecordBreak(GDBStub::BreakpointAddress bkpt);
+    void SaveContext(ThreadContext32& ctx) override {}
+    void SaveContext(ThreadContext64& ctx) override;
+    void LoadContext(const ThreadContext32& ctx) override {}
+    void LoadContext(const ThreadContext64& ctx) override;
 private:
    static void InterruptHook(uc_engine* uc, u32 int_no, void* user_data);
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 86e314c94..a82faf127 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -24,6 +24,7 @@
 #include "core/file_sys/sdmc_factory.h"
 #include "core/file_sys/vfs_concat.h"
 #include "core/file_sys/vfs_real.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hardware_interrupt_manager.h"
 #include "core/hle/kernel/client_port.h"
@@ -184,6 +185,8 @@ struct System::Impl {
    ResultStatus Load(System& system, Frontend::EmuWindow& emu_window,
                      const std::string& filepath) {
+        Core::Frontend::ScopeAcquireContext acquire_context{emu_window};
        app_loader = Loader::GetLoader(GetGameFileFromPath(virtual_filesystem, filepath));
        if (!app_loader) {
            LOG_CRITICAL(Core, "Failed to obtain loader for {}!", filepath);
diff --git a/src/core/core_manager.cpp b/src/core/core_manager.cpp
index 8eacf92dd..b6b797c80 100644
--- a/src/core/core_manager.cpp
+++ b/src/core/core_manager.cpp
@@ -6,9 +6,6 @@
 #include <mutex>
 #include "common/logging/log.h"
-#ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
-#endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
 #include "core/core.h"
diff --git a/src/core/frontend/emu_window.h b/src/core/frontend/emu_window.h
index 3376eedc5..5eb87fb63 100644
--- a/src/core/frontend/emu_window.h
+++ b/src/core/frontend/emu_window.h
@@ -26,9 +26,6 @@ public:
    /// Releases (dunno if this is the "right" word) the context from the caller thread
    virtual void DoneCurrent() = 0;
-    /// Swap buffers to display the next frame
-    virtual void SwapBuffers() = 0;
 };
 /**
diff --git a/src/core/frontend/framebuffer_layout.h b/src/core/frontend/framebuffer_layout.h
index 1d39c1faf..e9d0a40d3 100644
--- a/src/core/frontend/framebuffer_layout.h
+++ b/src/core/frontend/framebuffer_layout.h
@@ -29,6 +29,7 @@ enum class AspectRatio {
 struct FramebufferLayout {
    u32 width{ScreenUndocked::Width};
    u32 height{ScreenUndocked::Height};
+    bool is_srgb{};
    Common::Rectangle<u32> screen;
diff --git a/src/core/frontend/scope_acquire_context.cpp b/src/core/frontend/scope_acquire_context.cpp
new file mode 100644
index 000000000..878c3157c
--- /dev/null
+++ b/src/core/frontend/scope_acquire_context.cpp
@@ -0,0 +1,18 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#include "core/frontend/emu_window.h"
+#include "core/frontend/scope_acquire_context.h"
+namespace Core::Frontend {
+ScopeAcquireContext::ScopeAcquireContext(Core::Frontend::GraphicsContext& context)
+    : context{context} {
+    context.MakeCurrent();
+}
+ScopeAcquireContext::~ScopeAcquireContext() {
+    context.DoneCurrent();
+}
+} // namespace Core::Frontend
diff --git a/src/core/frontend/scope_acquire_window_context.h b/src/core/frontend/scope_acquire_context.h
index 2d9f6e825..7a65c0623 100644
--- a/src/core/frontend/scope_acquire_window_context.h
+++ b/src/core/frontend/scope_acquire_context.h
@@ -8,16 +8,16 @@
 namespace Core::Frontend {
-class EmuWindow;
+class GraphicsContext;
 /// Helper class to acquire/release window context within a given scope
-class ScopeAcquireWindowContext : NonCopyable {
+class ScopeAcquireContext : NonCopyable {
 public:
-    explicit ScopeAcquireWindowContext(Core::Frontend::EmuWindow& window);
+    explicit ScopeAcquireContext(Core::Frontend::GraphicsContext& context);
-    ~ScopeAcquireWindowContext();
+    ~ScopeAcquireContext();
 private:
-    Core::Frontend::EmuWindow& emu_window;
+    Core::Frontend::GraphicsContext& context;
 };
 } // namespace Core::Frontend
diff --git a/src/core/frontend/scope_acquire_window_context.cpp b/src/core/frontend/scope_acquire_window_context.cpp
deleted file mode 100644
index 3663dad17..000000000
--- a/src/core/frontend/scope_acquire_window_context.cpp
+++ /dev/null
@@ -1,18 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_window_context.h"
-namespace Core::Frontend {
-ScopeAcquireWindowContext::ScopeAcquireWindowContext(Core::Frontend::EmuWindow& emu_window_)
-    : emu_window{emu_window_} {
-    emu_window.MakeCurrent();
-}
-ScopeAcquireWindowContext::~ScopeAcquireWindowContext() {
-    emu_window.DoneCurrent();
-}
-} // namespace Core::Frontend
diff --git a/src/core/gdbstub/gdbstub.cpp b/src/core/gdbstub/gdbstub.cpp
index 67e95999d..e8d8871a7 100644
--- a/src/core/gdbstub/gdbstub.cpp
+++ b/src/core/gdbstub/gdbstub.cpp
@@ -217,7 +217,7 @@ static u64 RegRead(std::size_t id, Kernel::Thread* thread = nullptr) {
        return 0;
    }
-    const auto& thread_context = thread->GetContext();
+    const auto& thread_context = thread->GetContext64();
    if (id < SP_REGISTER) {
        return thread_context.cpu_registers[id];
@@ -239,7 +239,7 @@ static void RegWrite(std::size_t id, u64 val, Kernel::Thread* thread = nullptr)
        return;
    }
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
    if (id < SP_REGISTER) {
        thread_context.cpu_registers[id] = val;
@@ -259,7 +259,7 @@ static u128 FpuRead(std::size_t id, Kernel::Thread* thread = nullptr) {
        return u128{0};
    }
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
    if (id >= UC_ARM64_REG_Q0 && id < FPCR_REGISTER) {
        return thread_context.vector_registers[id - UC_ARM64_REG_Q0];
@@ -275,7 +275,7 @@ static void FpuWrite(std::size_t id, u128 val, Kernel::Thread* thread = nullptr)
        return;
    }
-    auto& thread_context = thread->GetContext();
+    auto& thread_context = thread->GetContext64();
    if (id >= UC_ARM64_REG_Q0 && id < FPCR_REGISTER) {
        thread_context.vector_registers[id - UC_ARM64_REG_Q0] = val;
@@ -916,7 +916,7 @@ static void WriteRegister() {
    // Update ARM context, skipping scheduler - no running threads at this point
    Core::System::GetInstance()
        .ArmInterface(current_core)
-        .LoadContext(current_thread->GetContext());
+        .LoadContext(current_thread->GetContext64());
    SendReply("OK");
 }
@@ -947,7 +947,7 @@ static void WriteRegisters() {
    // Update ARM context, skipping scheduler - no running threads at this point
    Core::System::GetInstance()
        .ArmInterface(current_core)
-        .LoadContext(current_thread->GetContext());
+        .LoadContext(current_thread->GetContext64());
    SendReply("OK");
 }
@@ -1019,7 +1019,7 @@ static void Step() {
        // Update ARM context, skipping scheduler - no running threads at this point
        Core::System::GetInstance()
            .ArmInterface(current_core)
-            .LoadContext(current_thread->GetContext());
+            .LoadContext(current_thread->GetContext64());
    }
    step_loop = true;
    halt_loop = true;
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 9232f4d7e..e47f1deed 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -186,6 +186,10 @@ struct KernelCore::Impl {
            return;
        }
+        for (auto& core : cores) {
+            core.SetIs64Bit(process->Is64BitProcess());
+        }
        system.Memory().SetCurrentPageTable(*process);
    }
diff --git a/src/core/hle/kernel/physical_core.cpp b/src/core/hle/kernel/physical_core.cpp
index 9303dd273..aa2787467 100644
--- a/src/core/hle/kernel/physical_core.cpp
+++ b/src/core/hle/kernel/physical_core.cpp
@@ -5,7 +5,8 @@
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
 #ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
+#include "core/arm/dynarmic/arm_dynarmic_32.h"
+#include "core/arm/dynarmic/arm_dynarmic_64.h"
 #endif
 #include "core/arm/exclusive_monitor.h"
 #include "core/arm/unicorn/arm_unicorn.h"
@@ -20,13 +21,17 @@ PhysicalCore::PhysicalCore(Core::System& system, std::size_t id,
                           Core::ExclusiveMonitor& exclusive_monitor)
    : core_index{id} {
 #ifdef ARCHITECTURE_x86_64
-    arm_interface = std::make_unique<Core::ARM_Dynarmic>(system, exclusive_monitor, core_index);
+    arm_interface_32 =
+        std::make_unique<Core::ARM_Dynarmic_32>(system, exclusive_monitor, core_index);
+    arm_interface_64 =
+        std::make_unique<Core::ARM_Dynarmic_64>(system, exclusive_monitor, core_index);
 #else
    arm_interface = std::make_shared<Core::ARM_Unicorn>(system);
    LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
 #endif
-    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface, core_index);
+    scheduler = std::make_unique<Kernel::Scheduler>(system, core_index);
 }
 PhysicalCore::~PhysicalCore() = default;
@@ -48,4 +53,12 @@ void PhysicalCore::Shutdown() {
    scheduler->Shutdown();
 }
+void PhysicalCore::SetIs64Bit(bool is_64_bit) {
+    if (is_64_bit) {
+        arm_interface = arm_interface_64.get();
+    } else {
+        arm_interface = arm_interface_32.get();
+    }
+}
 } // namespace Kernel
diff --git a/src/core/hle/kernel/physical_core.h b/src/core/hle/kernel/physical_core.h
index 4c32c0f1b..3269166be 100644
--- a/src/core/hle/kernel/physical_core.h
+++ b/src/core/hle/kernel/physical_core.h
@@ -68,10 +68,14 @@ public:
        return *scheduler;
    }
+    void SetIs64Bit(bool is_64_bit);
 private:
    std::size_t core_index;
-    std::unique_ptr<Core::ARM_Interface> arm_interface;
+    std::unique_ptr<Core::ARM_Interface> arm_interface_32;
+    std::unique_ptr<Core::ARM_Interface> arm_interface_64;
    std::unique_ptr<Kernel::Scheduler> scheduler;
+    Core::ARM_Interface* arm_interface{};
 };
 } // namespace Kernel
diff --git a/src/core/hle/kernel/process.cpp b/src/core/hle/kernel/process.cpp
index 2fcb7326c..edc414d69 100644
--- a/src/core/hle/kernel/process.cpp
+++ b/src/core/hle/kernel/process.cpp
@@ -42,7 +42,8 @@ void SetupMainThread(Process& owner_process, KernelCore& kernel, u32 priority) {
    // Register 1 must be a handle to the main thread
    const Handle thread_handle = owner_process.GetHandleTable().Create(thread).Unwrap();
-    thread->GetContext().cpu_registers[1] = thread_handle;
+    thread->GetContext32().cpu_registers[1] = thread_handle;
+    thread->GetContext64().cpu_registers[1] = thread_handle;
    // Threads by default are dormant, wake up the main thread so it runs when the scheduler fires
    thread->ResumeFromWait();
diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp
index c65f82fb7..1140c72a3 100644
--- a/src/core/hle/kernel/scheduler.cpp
+++ b/src/core/hle/kernel/scheduler.cpp
@@ -383,8 +383,8 @@ void GlobalScheduler::Unlock() {
    // TODO(Blinkhawk): Setup the interrupts and change context on current core.
 }
-Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, std::size_t core_id)
+Scheduler::Scheduler(Core::System& system, std::size_t core_id)
-    : system(system), cpu_core(cpu_core), core_id(core_id) {}
+    : system{system}, core_id{core_id} {}
 Scheduler::~Scheduler() = default;
@@ -422,9 +422,10 @@ void Scheduler::UnloadThread() {
    // Save context for previous thread
    if (previous_thread) {
-        cpu_core.SaveContext(previous_thread->GetContext());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext32());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext64());
        // Save the TPIDR_EL0 system register in case it was modified.
-        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+        previous_thread->SetTPIDR_EL0(system.ArmInterface(core_id).GetTPIDR_EL0());
        if (previous_thread->GetStatus() == ThreadStatus::Running) {
            // This is only the case when a reschedule is triggered without the current thread
@@ -451,9 +452,10 @@ void Scheduler::SwitchContext() {
    // Save context for previous thread
    if (previous_thread) {
-        cpu_core.SaveContext(previous_thread->GetContext());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext32());
+        system.ArmInterface(core_id).SaveContext(previous_thread->GetContext64());
        // Save the TPIDR_EL0 system register in case it was modified.
-        previous_thread->SetTPIDR_EL0(cpu_core.GetTPIDR_EL0());
+        previous_thread->SetTPIDR_EL0(system.ArmInterface(core_id).GetTPIDR_EL0());
        if (previous_thread->GetStatus() == ThreadStatus::Running) {
            // This is only the case when a reschedule is triggered without the current thread
@@ -481,9 +483,10 @@ void Scheduler::SwitchContext() {
            system.Kernel().MakeCurrentProcess(thread_owner_process);
        }
-        cpu_core.LoadContext(new_thread->GetContext());
+        system.ArmInterface(core_id).LoadContext(new_thread->GetContext32());
-        cpu_core.SetTlsAddress(new_thread->GetTLSAddress());
+        system.ArmInterface(core_id).LoadContext(new_thread->GetContext64());
-        cpu_core.SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
+        system.ArmInterface(core_id).SetTlsAddress(new_thread->GetTLSAddress());
+        system.ArmInterface(core_id).SetTPIDR_EL0(new_thread->GetTPIDR_EL0());
    } else {
        current_thread = nullptr;
        // Note: We do not reset the current process and current page table when idling because
diff --git a/src/core/hle/kernel/scheduler.h b/src/core/hle/kernel/scheduler.h
index 1c93a838c..07df33f9c 100644
--- a/src/core/hle/kernel/scheduler.h
+++ b/src/core/hle/kernel/scheduler.h
@@ -181,7 +181,7 @@ private:
 class Scheduler final {
 public:
-    explicit Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, std::size_t core_id);
+    explicit Scheduler(Core::System& system, std::size_t core_id);
    ~Scheduler();
    /// Returns whether there are any threads that are ready to run.
@@ -235,7 +235,6 @@ private:
    std::shared_ptr<Thread> selected_thread = nullptr;
    Core::System& system;
-    Core::ARM_Interface& cpu_core;
    u64 last_context_switch_time = 0;
    u64 idle_selection_count = 0;
    const std::size_t core_id;
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index fd91779a3..4ffc113c2 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -187,6 +187,13 @@ static ResultCode SetHeapSize(Core::System& system, VAddr* heap_addr, u64 heap_s
    return RESULT_SUCCESS;
 }
+static ResultCode SetHeapSize32(Core::System& system, u32* heap_addr, u32 heap_size) {
+    VAddr temp_heap_addr{};
+    const ResultCode result{SetHeapSize(system, &temp_heap_addr, heap_size)};
+    *heap_addr = static_cast<u32>(temp_heap_addr);
+    return result;
+}
 static ResultCode SetMemoryPermission(Core::System& system, VAddr addr, u64 size, u32 prot) {
    LOG_TRACE(Kernel_SVC, "called, addr=0x{:X}, size=0x{:X}, prot=0x{:X}", addr, size, prot);
@@ -371,6 +378,12 @@ static ResultCode ConnectToNamedPort(Core::System& system, Handle* out_handle,
    return RESULT_SUCCESS;
 }
+static ResultCode ConnectToNamedPort32(Core::System& system, Handle* out_handle,
+                                       u32 port_name_address) {
+    return ConnectToNamedPort(system, out_handle, port_name_address);
+}
 /// Makes a blocking IPC call to an OS service.
 static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
    const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
@@ -390,6 +403,10 @@ static ResultCode SendSyncRequest(Core::System& system, Handle handle) {
    return session->SendSyncRequest(SharedFrom(thread), system.Memory());
 }
+static ResultCode SendSyncRequest32(Core::System& system, Handle handle) {
+    return SendSyncRequest(system, handle);
+}
 /// Get the ID for the specified thread.
 static ResultCode GetThreadId(Core::System& system, u64* thread_id, Handle thread_handle) {
    LOG_TRACE(Kernel_SVC, "called thread=0x{:08X}", thread_handle);
@@ -405,6 +422,17 @@ static ResultCode GetThreadId(Core::System& system, u64* thread_id, Handle threa
    return RESULT_SUCCESS;
 }
+static ResultCode GetThreadId32(Core::System& system, u32* thread_id_low, u32* thread_id_high,
+                                Handle thread_handle) {
+    u64 thread_id{};
+    const ResultCode result{GetThreadId(system, &thread_id, thread_handle)};
+    *thread_id_low = static_cast<u32>(thread_id >> 32);
+    *thread_id_high = static_cast<u32>(thread_id & std::numeric_limits<u32>::max());
+    return result;
+}
 /// Gets the ID of the specified process or a specified thread's owning process.
 static ResultCode GetProcessId(Core::System& system, u64* process_id, Handle handle) {
    LOG_DEBUG(Kernel_SVC, "called handle=0x{:08X}", handle);
@@ -479,6 +507,12 @@ static ResultCode WaitSynchronization(Core::System& system, Handle* index, VAddr
    return result;
 }
+static ResultCode WaitSynchronization32(Core::System& system, u32 timeout_low, u32 handles_address,
+                                        s32 handle_count, u32 timeout_high, Handle* index) {
+    const s64 nano_seconds{(static_cast<s64>(timeout_high) << 32) | static_cast<s64>(timeout_low)};
+    return WaitSynchronization(system, index, handles_address, handle_count, nano_seconds);
+}
 /// Resumes a thread waiting on WaitSynchronization
 static ResultCode CancelSynchronization(Core::System& system, Handle thread_handle) {
    LOG_TRACE(Kernel_SVC, "called thread=0x{:X}", thread_handle);
@@ -917,6 +951,18 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, u64 ha
    }
 }
+static ResultCode GetInfo32(Core::System& system, u32* result_low, u32* result_high, u32 sub_id_low,
+                            u32 info_id, u32 handle, u32 sub_id_high) {
+    const u64 sub_id{static_cast<u64>(sub_id_low | (static_cast<u64>(sub_id_high) << 32))};
+    u64 res_value{};
+    const ResultCode result{GetInfo(system, &res_value, info_id, handle, sub_id)};
+    *result_high = static_cast<u32>(res_value >> 32);
+    *result_low = static_cast<u32>(res_value & std::numeric_limits<u32>::max());
+    return result;
+}
 /// Maps memory at a desired address
 static ResultCode MapPhysicalMemory(Core::System& system, VAddr addr, u64 size) {
    LOG_DEBUG(Kernel_SVC, "called, addr=0x{:016X}, size=0x{:X}", addr, size);
@@ -1058,7 +1104,7 @@ static ResultCode GetThreadContext(Core::System& system, VAddr thread_context, H
        return ERR_BUSY;
    }
-    Core::ARM_Interface::ThreadContext ctx = thread->GetContext();
+    Core::ARM_Interface::ThreadContext64 ctx = thread->GetContext64();
    // Mask away mode bits, interrupt bits, IL bit, and other reserved bits.
    ctx.pstate &= 0xFF0FFE20;
@@ -1088,6 +1134,10 @@ static ResultCode GetThreadPriority(Core::System& system, u32* priority, Handle
    return RESULT_SUCCESS;
 }
+static ResultCode GetThreadPriority32(Core::System& system, u32* priority, Handle handle) {
+    return GetThreadPriority(system, priority, handle);
+}
 /// Sets the priority for the specified thread
 static ResultCode SetThreadPriority(Core::System& system, Handle handle, u32 priority) {
    LOG_TRACE(Kernel_SVC, "called");
@@ -1259,6 +1309,11 @@ static ResultCode QueryMemory(Core::System& system, VAddr memory_info_address,
                              query_address);
 }
+static ResultCode QueryMemory32(Core::System& system, u32 memory_info_address,
+                                u32 page_info_address, u32 query_address) {
+    return QueryMemory(system, memory_info_address, page_info_address, query_address);
+}
 static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_handle, u64 dst_address,
                                       u64 src_address, u64 size) {
    LOG_DEBUG(Kernel_SVC,
@@ -1675,6 +1730,10 @@ static void SignalProcessWideKey(Core::System& system, VAddr condition_variable_
    }
 }
+static void SignalProcessWideKey32(Core::System& system, u32 condition_variable_addr, s32 target) {
+    SignalProcessWideKey(system, condition_variable_addr, target);
+}
 // Wait for an address (via Address Arbiter)
 static ResultCode WaitForAddress(Core::System& system, VAddr address, u32 type, s32 value,
                                 s64 timeout) {
@@ -1760,6 +1819,10 @@ static ResultCode CloseHandle(Core::System& system, Handle handle) {
    return handle_table.Close(handle);
 }
+static ResultCode CloseHandle32(Core::System& system, Handle handle) {
+    return CloseHandle(system, handle);
+}
 /// Clears the signaled state of an event or process.
 static ResultCode ResetSignal(Core::System& system, Handle handle) {
    LOG_DEBUG(Kernel_SVC, "called handle 0x{:08X}", handle);
@@ -2317,69 +2380,196 @@ struct FunctionDef {
 };
 } // namespace
-static const FunctionDef SVC_Table[] = {
+static const FunctionDef SVC_Table_32[] = {
    {0x00, nullptr, "Unknown"},
-    {0x01, SvcWrap<SetHeapSize>, "SetHeapSize"},
+    {0x01, SvcWrap32<SetHeapSize32>, "SetHeapSize32"},
-    {0x02, SvcWrap<SetMemoryPermission>, "SetMemoryPermission"},
+    {0x02, nullptr, "Unknown"},
-    {0x03, SvcWrap<SetMemoryAttribute>, "SetMemoryAttribute"},
+    {0x03, nullptr, "SetMemoryAttribute32"},
-    {0x04, SvcWrap<MapMemory>, "MapMemory"},
+    {0x04, nullptr, "MapMemory32"},
-    {0x05, SvcWrap<UnmapMemory>, "UnmapMemory"},
+    {0x05, nullptr, "UnmapMemory32"},
-    {0x06, SvcWrap<QueryMemory>, "QueryMemory"},
+    {0x06, SvcWrap32<QueryMemory32>, "QueryMemory32"},
-    {0x07, SvcWrap<ExitProcess>, "ExitProcess"},
+    {0x07, nullptr, "ExitProcess32"},
-    {0x08, SvcWrap<CreateThread>, "CreateThread"},
+    {0x08, nullptr, "CreateThread32"},
-    {0x09, SvcWrap<StartThread>, "StartThread"},
+    {0x09, nullptr, "StartThread32"},
-    {0x0A, SvcWrap<ExitThread>, "ExitThread"},
+    {0x0a, nullptr, "ExitThread32"},
-    {0x0B, SvcWrap<SleepThread>, "SleepThread"},
+    {0x0b, nullptr, "SleepThread32"},
-    {0x0C, SvcWrap<GetThreadPriority>, "GetThreadPriority"},
+    {0x0c, SvcWrap32<GetThreadPriority32>, "GetThreadPriority32"},
-    {0x0D, SvcWrap<SetThreadPriority>, "SetThreadPriority"},
+    {0x0d, nullptr, "SetThreadPriority32"},
-    {0x0E, SvcWrap<GetThreadCoreMask>, "GetThreadCoreMask"},
+    {0x0e, nullptr, "GetThreadCoreMask32"},
-    {0x0F, SvcWrap<SetThreadCoreMask>, "SetThreadCoreMask"},
+    {0x0f, nullptr, "SetThreadCoreMask32"},
-    {0x10, SvcWrap<GetCurrentProcessorNumber>, "GetCurrentProcessorNumber"},
+    {0x10, nullptr, "GetCurrentProcessorNumber32"},
-    {0x11, SvcWrap<SignalEvent>, "SignalEvent"},
+    {0x11, nullptr, "SignalEvent32"},
-    {0x12, SvcWrap<ClearEvent>, "ClearEvent"},
+    {0x12, nullptr, "ClearEvent32"},
-    {0x13, SvcWrap<MapSharedMemory>, "MapSharedMemory"},
+    {0x13, nullptr, "MapSharedMemory32"},
-    {0x14, SvcWrap<UnmapSharedMemory>, "UnmapSharedMemory"},
+    {0x14, nullptr, "UnmapSharedMemory32"},
-    {0x15, SvcWrap<CreateTransferMemory>, "CreateTransferMemory"},
+    {0x15, nullptr, "CreateTransferMemory32"},
-    {0x16, SvcWrap<CloseHandle>, "CloseHandle"},
+    {0x16, SvcWrap32<CloseHandle32>, "CloseHandle32"},
-    {0x17, SvcWrap<ResetSignal>, "ResetSignal"},
+    {0x17, nullptr, "ResetSignal32"},
-    {0x18, SvcWrap<WaitSynchronization>, "WaitSynchronization"},
+    {0x18, SvcWrap32<WaitSynchronization32>, "WaitSynchronization32"},
-    {0x19, SvcWrap<CancelSynchronization>, "CancelSynchronization"},
+    {0x19, nullptr, "CancelSynchronization32"},
-    {0x1A, SvcWrap<ArbitrateLock>, "ArbitrateLock"},
+    {0x1a, nullptr, "ArbitrateLock32"},
-    {0x1B, SvcWrap<ArbitrateUnlock>, "ArbitrateUnlock"},
+    {0x1b, nullptr, "ArbitrateUnlock32"},
-    {0x1C, SvcWrap<WaitProcessWideKeyAtomic>, "WaitProcessWideKeyAtomic"},
+    {0x1c, nullptr, "WaitProcessWideKeyAtomic32"},
-    {0x1D, SvcWrap<SignalProcessWideKey>, "SignalProcessWideKey"},
+    {0x1d, SvcWrap32<SignalProcessWideKey32>, "SignalProcessWideKey32"},
-    {0x1E, SvcWrap<GetSystemTick>, "GetSystemTick"},
+    {0x1e, nullptr, "GetSystemTick32"},
-    {0x1F, SvcWrap<ConnectToNamedPort>, "ConnectToNamedPort"},
+    {0x1f, SvcWrap32<ConnectToNamedPort32>, "ConnectToNamedPort32"},
+    {0x20, nullptr, "Unknown"},
+    {0x21, SvcWrap32<SendSyncRequest32>, "SendSyncRequest32"},
+    {0x22, nullptr, "SendSyncRequestWithUserBuffer32"},
+    {0x23, nullptr, "Unknown"},
+    {0x24, nullptr, "GetProcessId32"},
+    {0x25, SvcWrap32<GetThreadId32>, "GetThreadId32"},
+    {0x26, nullptr, "Break32"},
+    {0x27, nullptr, "OutputDebugString32"},
+    {0x28, nullptr, "Unknown"},
+    {0x29, SvcWrap32<GetInfo32>, "GetInfo32"},
+    {0x2a, nullptr, "Unknown"},
+    {0x2b, nullptr, "Unknown"},
+    {0x2c, nullptr, "MapPhysicalMemory32"},
+    {0x2d, nullptr, "UnmapPhysicalMemory32"},
+    {0x2e, nullptr, "Unknown"},
+    {0x2f, nullptr, "Unknown"},
+    {0x30, nullptr, "Unknown"},
+    {0x31, nullptr, "Unknown"},
+    {0x32, nullptr, "SetThreadActivity32"},
+    {0x33, nullptr, "GetThreadContext32"},
+    {0x34, nullptr, "WaitForAddress32"},
+    {0x35, nullptr, "SignalToAddress32"},
+    {0x36, nullptr, "Unknown"},
+    {0x37, nullptr, "Unknown"},
+    {0x38, nullptr, "Unknown"},
+    {0x39, nullptr, "Unknown"},
+    {0x3a, nullptr, "Unknown"},
+    {0x3b, nullptr, "Unknown"},
+    {0x3c, nullptr, "Unknown"},
+    {0x3d, nullptr, "Unknown"},
+    {0x3e, nullptr, "Unknown"},
+    {0x3f, nullptr, "Unknown"},
+    {0x40, nullptr, "CreateSession32"},
+    {0x41, nullptr, "AcceptSession32"},
+    {0x42, nullptr, "Unknown"},
+    {0x43, nullptr, "ReplyAndReceive32"},
+    {0x44, nullptr, "Unknown"},
+    {0x45, nullptr, "CreateEvent32"},
+    {0x46, nullptr, "Unknown"},
+    {0x47, nullptr, "Unknown"},
+    {0x48, nullptr, "Unknown"},
+    {0x49, nullptr, "Unknown"},
+    {0x4a, nullptr, "Unknown"},
+    {0x4b, nullptr, "Unknown"},
+    {0x4c, nullptr, "Unknown"},
+    {0x4d, nullptr, "Unknown"},
+    {0x4e, nullptr, "Unknown"},
+    {0x4f, nullptr, "Unknown"},
+    {0x50, nullptr, "Unknown"},
+    {0x51, nullptr, "Unknown"},
+    {0x52, nullptr, "Unknown"},
+    {0x53, nullptr, "Unknown"},
+    {0x54, nullptr, "Unknown"},
+    {0x55, nullptr, "Unknown"},
+    {0x56, nullptr, "Unknown"},
+    {0x57, nullptr, "Unknown"},
+    {0x58, nullptr, "Unknown"},
+    {0x59, nullptr, "Unknown"},
+    {0x5a, nullptr, "Unknown"},
+    {0x5b, nullptr, "Unknown"},
+    {0x5c, nullptr, "Unknown"},
+    {0x5d, nullptr, "Unknown"},
+    {0x5e, nullptr, "Unknown"},
+    {0x5F, nullptr, "FlushProcessDataCache32"},
+    {0x60, nullptr, "Unknown"},
+    {0x61, nullptr, "Unknown"},
+    {0x62, nullptr, "Unknown"},
+    {0x63, nullptr, "Unknown"},
+    {0x64, nullptr, "Unknown"},
+    {0x65, nullptr, "GetProcessList32"},
+    {0x66, nullptr, "Unknown"},
+    {0x67, nullptr, "Unknown"},
+    {0x68, nullptr, "Unknown"},
+    {0x69, nullptr, "Unknown"},
+    {0x6A, nullptr, "Unknown"},
+    {0x6B, nullptr, "Unknown"},
+    {0x6C, nullptr, "Unknown"},
+    {0x6D, nullptr, "Unknown"},
+    {0x6E, nullptr, "Unknown"},
+    {0x6f, nullptr, "GetSystemInfo32"},
+    {0x70, nullptr, "CreatePort32"},
+    {0x71, nullptr, "ManageNamedPort32"},
+    {0x72, nullptr, "ConnectToPort32"},
+    {0x73, nullptr, "SetProcessMemoryPermission32"},
+    {0x74, nullptr, "Unknown"},
+    {0x75, nullptr, "Unknown"},
+    {0x76, nullptr, "Unknown"},
+    {0x77, nullptr, "MapProcessCodeMemory32"},
+    {0x78, nullptr, "UnmapProcessCodeMemory32"},
+    {0x79, nullptr, "Unknown"},
+    {0x7A, nullptr, "Unknown"},
+    {0x7B, nullptr, "TerminateProcess32"},
+};
+static const FunctionDef SVC_Table_64[] = {
+    {0x00, nullptr, "Unknown"},
+    {0x01, SvcWrap64<SetHeapSize>, "SetHeapSize"},
+    {0x02, SvcWrap64<SetMemoryPermission>, "SetMemoryPermission"},
+    {0x03, SvcWrap64<SetMemoryAttribute>, "SetMemoryAttribute"},
+    {0x04, SvcWrap64<MapMemory>, "MapMemory"},
+    {0x05, SvcWrap64<UnmapMemory>, "UnmapMemory"},
+    {0x06, SvcWrap64<QueryMemory>, "QueryMemory"},
+    {0x07, SvcWrap64<ExitProcess>, "ExitProcess"},
+    {0x08, SvcWrap64<CreateThread>, "CreateThread"},
+    {0x09, SvcWrap64<StartThread>, "StartThread"},
+    {0x0A, SvcWrap64<ExitThread>, "ExitThread"},
+    {0x0B, SvcWrap64<SleepThread>, "SleepThread"},
+    {0x0C, SvcWrap64<GetThreadPriority>, "GetThreadPriority"},
+    {0x0D, SvcWrap64<SetThreadPriority>, "SetThreadPriority"},
+    {0x0E, SvcWrap64<GetThreadCoreMask>, "GetThreadCoreMask"},
+    {0x0F, SvcWrap64<SetThreadCoreMask>, "SetThreadCoreMask"},
+    {0x10, SvcWrap64<GetCurrentProcessorNumber>, "GetCurrentProcessorNumber"},
+    {0x11, SvcWrap64<SignalEvent>, "SignalEvent"},
+    {0x12, SvcWrap64<ClearEvent>, "ClearEvent"},
+    {0x13, SvcWrap64<MapSharedMemory>, "MapSharedMemory"},
+    {0x14, SvcWrap64<UnmapSharedMemory>, "UnmapSharedMemory"},
+    {0x15, SvcWrap64<CreateTransferMemory>, "CreateTransferMemory"},
+    {0x16, SvcWrap64<CloseHandle>, "CloseHandle"},
+    {0x17, SvcWrap64<ResetSignal>, "ResetSignal"},
+    {0x18, SvcWrap64<WaitSynchronization>, "WaitSynchronization"},
+    {0x19, SvcWrap64<CancelSynchronization>, "CancelSynchronization"},
+    {0x1A, SvcWrap64<ArbitrateLock>, "ArbitrateLock"},
+    {0x1B, SvcWrap64<ArbitrateUnlock>, "ArbitrateUnlock"},
+    {0x1C, SvcWrap64<WaitProcessWideKeyAtomic>, "WaitProcessWideKeyAtomic"},
+    {0x1D, SvcWrap64<SignalProcessWideKey>, "SignalProcessWideKey"},
+    {0x1E, SvcWrap64<GetSystemTick>, "GetSystemTick"},
+    {0x1F, SvcWrap64<ConnectToNamedPort>, "ConnectToNamedPort"},
    {0x20, nullptr, "SendSyncRequestLight"},
-    {0x21, SvcWrap<SendSyncRequest>, "SendSyncRequest"},
+    {0x21, SvcWrap64<SendSyncRequest>, "SendSyncRequest"},
    {0x22, nullptr, "SendSyncRequestWithUserBuffer"},
    {0x23, nullptr, "SendAsyncRequestWithUserBuffer"},
-    {0x24, SvcWrap<GetProcessId>, "GetProcessId"},
+    {0x24, SvcWrap64<GetProcessId>, "GetProcessId"},
-    {0x25, SvcWrap<GetThreadId>, "GetThreadId"},
+    {0x25, SvcWrap64<GetThreadId>, "GetThreadId"},
-    {0x26, SvcWrap<Break>, "Break"},
+    {0x26, SvcWrap64<Break>, "Break"},
-    {0x27, SvcWrap<OutputDebugString>, "OutputDebugString"},
+    {0x27, SvcWrap64<OutputDebugString>, "OutputDebugString"},
    {0x28, nullptr, "ReturnFromException"},
-    {0x29, SvcWrap<GetInfo>, "GetInfo"},
+    {0x29, SvcWrap64<GetInfo>, "GetInfo"},
    {0x2A, nullptr, "FlushEntireDataCache"},
    {0x2B, nullptr, "FlushDataCache"},
-    {0x2C, SvcWrap<MapPhysicalMemory>, "MapPhysicalMemory"},
+    {0x2C, SvcWrap64<MapPhysicalMemory>, "MapPhysicalMemory"},
-    {0x2D, SvcWrap<UnmapPhysicalMemory>, "UnmapPhysicalMemory"},
+    {0x2D, SvcWrap64<UnmapPhysicalMemory>, "UnmapPhysicalMemory"},
    {0x2E, nullptr, "GetFutureThreadInfo"},
    {0x2F, nullptr, "GetLastThreadInfo"},
-    {0x30, SvcWrap<GetResourceLimitLimitValue>, "GetResourceLimitLimitValue"},
+    {0x30, SvcWrap64<GetResourceLimitLimitValue>, "GetResourceLimitLimitValue"},
-    {0x31, SvcWrap<GetResourceLimitCurrentValue>, "GetResourceLimitCurrentValue"},
+    {0x31, SvcWrap64<GetResourceLimitCurrentValue>, "GetResourceLimitCurrentValue"},
-    {0x32, SvcWrap<SetThreadActivity>, "SetThreadActivity"},
+    {0x32, SvcWrap64<SetThreadActivity>, "SetThreadActivity"},
-    {0x33, SvcWrap<GetThreadContext>, "GetThreadContext"},
+    {0x33, SvcWrap64<GetThreadContext>, "GetThreadContext"},
-    {0x34, SvcWrap<WaitForAddress>, "WaitForAddress"},
+    {0x34, SvcWrap64<WaitForAddress>, "WaitForAddress"},
-    {0x35, SvcWrap<SignalToAddress>, "SignalToAddress"},
+    {0x35, SvcWrap64<SignalToAddress>, "SignalToAddress"},
    {0x36, nullptr, "SynchronizePreemptionState"},
    {0x37, nullptr, "Unknown"},
    {0x38, nullptr, "Unknown"},
    {0x39, nullptr, "Unknown"},
    {0x3A, nullptr, "Unknown"},
    {0x3B, nullptr, "Unknown"},
-    {0x3C, SvcWrap<KernelDebug>, "KernelDebug"},
+    {0x3C, SvcWrap64<KernelDebug>, "KernelDebug"},
-    {0x3D, SvcWrap<ChangeKernelTraceState>, "ChangeKernelTraceState"},
+    {0x3D, SvcWrap64<ChangeKernelTraceState>, "ChangeKernelTraceState"},
    {0x3E, nullptr, "Unknown"},
    {0x3F, nullptr, "Unknown"},
    {0x40, nullptr, "CreateSession"},
@@ -2387,7 +2577,7 @@ static const FunctionDef SVC_Table[] = {
    {0x42, nullptr, "ReplyAndReceiveLight"},
    {0x43, nullptr, "ReplyAndReceive"},
    {0x44, nullptr, "ReplyAndReceiveWithUserBuffer"},
-    {0x45, SvcWrap<CreateEvent>, "CreateEvent"},
+    {0x45, SvcWrap64<CreateEvent>, "CreateEvent"},
    {0x46, nullptr, "Unknown"},
    {0x47, nullptr, "Unknown"},
    {0x48, nullptr, "MapPhysicalMemoryUnsafe"},
@@ -2398,9 +2588,9 @@ static const FunctionDef SVC_Table[] = {
    {0x4D, nullptr, "SleepSystem"},
    {0x4E, nullptr, "ReadWriteRegister"},
    {0x4F, nullptr, "SetProcessActivity"},
-    {0x50, SvcWrap<CreateSharedMemory>, "CreateSharedMemory"},
+    {0x50, SvcWrap64<CreateSharedMemory>, "CreateSharedMemory"},
-    {0x51, SvcWrap<MapTransferMemory>, "MapTransferMemory"},
+    {0x51, SvcWrap64<MapTransferMemory>, "MapTransferMemory"},
-    {0x52, SvcWrap<UnmapTransferMemory>, "UnmapTransferMemory"},
+    {0x52, SvcWrap64<UnmapTransferMemory>, "UnmapTransferMemory"},
    {0x53, nullptr, "CreateInterruptEvent"},
    {0x54, nullptr, "QueryPhysicalAddress"},
    {0x55, nullptr, "QueryIoMapping"},
@@ -2419,8 +2609,8 @@ static const FunctionDef SVC_Table[] = {
    {0x62, nullptr, "TerminateDebugProcess"},
    {0x63, nullptr, "GetDebugEvent"},
    {0x64, nullptr, "ContinueDebugEvent"},
-    {0x65, SvcWrap<GetProcessList>, "GetProcessList"},
+    {0x65, SvcWrap64<GetProcessList>, "GetProcessList"},
-    {0x66, SvcWrap<GetThreadList>, "GetThreadList"},
+    {0x66, SvcWrap64<GetThreadList>, "GetThreadList"},
    {0x67, nullptr, "GetDebugThreadContext"},
    {0x68, nullptr, "SetDebugThreadContext"},
    {0x69, nullptr, "QueryDebugProcessMemory"},
@@ -2436,24 +2626,32 @@ static const FunctionDef SVC_Table[] = {
    {0x73, nullptr, "SetProcessMemoryPermission"},
    {0x74, nullptr, "MapProcessMemory"},
    {0x75, nullptr, "UnmapProcessMemory"},
-    {0x76, SvcWrap<QueryProcessMemory>, "QueryProcessMemory"},
+    {0x76, SvcWrap64<QueryProcessMemory>, "QueryProcessMemory"},
-    {0x77, SvcWrap<MapProcessCodeMemory>, "MapProcessCodeMemory"},
+    {0x77, SvcWrap64<MapProcessCodeMemory>, "MapProcessCodeMemory"},
-    {0x78, SvcWrap<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
+    {0x78, SvcWrap64<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
    {0x79, nullptr, "CreateProcess"},
    {0x7A, nullptr, "StartProcess"},
    {0x7B, nullptr, "TerminateProcess"},
-    {0x7C, SvcWrap<GetProcessInfo>, "GetProcessInfo"},
+    {0x7C, SvcWrap64<GetProcessInfo>, "GetProcessInfo"},
-    {0x7D, SvcWrap<CreateResourceLimit>, "CreateResourceLimit"},
+    {0x7D, SvcWrap64<CreateResourceLimit>, "CreateResourceLimit"},
-    {0x7E, SvcWrap<SetResourceLimitLimitValue>, "SetResourceLimitLimitValue"},
+    {0x7E, SvcWrap64<SetResourceLimitLimitValue>, "SetResourceLimitLimitValue"},
    {0x7F, nullptr, "CallSecureMonitor"},
 };
-static const FunctionDef* GetSVCInfo(u32 func_num) {
+static const FunctionDef* GetSVCInfo32(u32 func_num) {
-    if (func_num >= std::size(SVC_Table)) {
+    if (func_num >= std::size(SVC_Table_32)) {
+        LOG_ERROR(Kernel_SVC, "Unknown svc=0x{:02X}", func_num);
+        return nullptr;
+    }
+    return &SVC_Table_32[func_num];
+}
+static const FunctionDef* GetSVCInfo64(u32 func_num) {
+    if (func_num >= std::size(SVC_Table_64)) {
        LOG_ERROR(Kernel_SVC, "Unknown svc=0x{:02X}", func_num);
        return nullptr;
    }
-    return &SVC_Table[func_num];
+    return &SVC_Table_64[func_num];
 }
 MICROPROFILE_DEFINE(Kernel_SVC, "Kernel", "SVC", MP_RGB(70, 200, 70));
@@ -2464,7 +2662,8 @@ void CallSVC(Core::System& system, u32 immediate) {
    // Lock the global kernel mutex when we enter the kernel HLE.
    std::lock_guard lock{HLE::g_hle_lock};
-    const FunctionDef* info = GetSVCInfo(immediate);
+    const FunctionDef* info = system.CurrentProcess()->Is64BitProcess() ? GetSVCInfo64(immediate)
+                                                                        : GetSVCInfo32(immediate);
    if (info) {
        if (info->func) {
            info->func(system);
diff --git a/src/core/hle/kernel/svc_wrap.h b/src/core/hle/kernel/svc_wrap.h
index 29a2cfa9d..7d735e3fa 100644
--- a/src/core/hle/kernel/svc_wrap.h
+++ b/src/core/hle/kernel/svc_wrap.h
@@ -15,6 +15,10 @@ static inline u64 Param(const Core::System& system, int n) {
    return system.CurrentArmInterface().GetReg(n);
 }
+static inline u32 Param32(const Core::System& system, int n) {
+    return static_cast<u32>(system.CurrentArmInterface().GetReg(n));
+}
 /**
 * HLE a function return from the current ARM userland process
 * @param system System context
@@ -24,40 +28,44 @@ static inline void FuncReturn(Core::System& system, u64 result) {
    system.CurrentArmInterface().SetReg(0, result);
 }
+static inline void FuncReturn32(Core::System& system, u32 result) {
+    system.CurrentArmInterface().SetReg(0, (u64)result);
+}
 ////////////////////////////////////////////////////////////////////////////////////////////////////
 // Function wrappers that return type ResultCode
 template <ResultCode func(Core::System&, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0)).raw);
 }
 template <ResultCode func(Core::System&, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), Param(system, 1)).raw);
 }
 template <ResultCode func(Core::System&, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
 }
 template <ResultCode func(Core::System&, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(
        system,
        func(system, static_cast<u32>(Param(system, 0)), static_cast<u32>(Param(system, 1))).raw);
 }
 template <ResultCode func(Core::System&, u32, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                            Param(system, 2), Param(system, 3))
                           .raw);
 }
 template <ResultCode func(Core::System&, u32*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param = 0;
    const u32 retval = func(system, &param).raw;
    system.CurrentArmInterface().SetReg(1, param);
@@ -65,7 +73,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u32*, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
    system.CurrentArmInterface().SetReg(1, param_1);
@@ -73,7 +81,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u32*, u32*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    u32 param_2 = 0;
    const u32 retval = func(system, &param_1, &param_2).raw;
@@ -86,7 +94,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u32*, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    const u32 retval = func(system, &param_1, Param(system, 1)).raw;
    system.CurrentArmInterface().SetReg(1, param_1);
@@ -94,7 +102,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u32*, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    const u32 retval =
        func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2))).raw;
@@ -104,7 +112,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u64*, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u64 param_1 = 0;
    const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1))).raw;
@@ -113,12 +121,12 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1))).raw);
 }
 template <ResultCode func(Core::System&, u64*, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u64 param_1 = 0;
    const u32 retval = func(system, &param_1, Param(system, 1)).raw;
@@ -127,7 +135,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u64*, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u64 param_1 = 0;
    const u32 retval = func(system, &param_1, static_cast<u32>(Param(system, 1)),
                            static_cast<u32>(Param(system, 2)))
@@ -138,19 +146,19 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1)).raw);
 }
 template <ResultCode func(Core::System&, u32, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)),
                            static_cast<u32>(Param(system, 1)), Param(system, 2))
                           .raw);
 }
 template <ResultCode func(Core::System&, u32, u32*, u64*)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    u64 param_2 = 0;
    const ResultCode retval = func(system, static_cast<u32>(Param(system, 2)), &param_1, &param_2);
@@ -161,54 +169,54 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u64, u64, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                            static_cast<u32>(Param(system, 2)), static_cast<u32>(Param(system, 3)))
                           .raw);
 }
 template <ResultCode func(Core::System&, u64, u64, u32, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                            static_cast<u32>(Param(system, 2)), Param(system, 3))
                           .raw);
 }
 template <ResultCode func(Core::System&, u32, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                            static_cast<u32>(Param(system, 2)))
                           .raw);
 }
 template <ResultCode func(Core::System&, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), Param(system, 1), Param(system, 2)).raw);
 }
 template <ResultCode func(Core::System&, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(
        system,
        func(system, Param(system, 0), Param(system, 1), static_cast<u32>(Param(system, 2))).raw);
 }
 template <ResultCode func(Core::System&, u32, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0)), Param(system, 1),
                            Param(system, 2), static_cast<u32>(Param(system, 3)))
                           .raw);
 }
 template <ResultCode func(Core::System&, u32, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(
        system,
        func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2)).raw);
 }
 template <ResultCode func(Core::System&, u32*, u64, u64, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
                            static_cast<s64>(Param(system, 3)))
@@ -219,14 +227,14 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u64, u64, u32, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), Param(system, 1),
                            static_cast<u32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
                           .raw);
 }
 template <ResultCode func(Core::System&, u64*, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u64 param_1 = 0;
    const u32 retval =
        func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3)).raw;
@@ -236,7 +244,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u32*, u64, u64, u64, u32, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2), Param(system, 3),
                            static_cast<u32>(Param(system, 4)), static_cast<s32>(Param(system, 5)))
@@ -247,7 +255,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u32*, u64, u64, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    const u32 retval = func(system, &param_1, Param(system, 1), Param(system, 2),
                            static_cast<u32>(Param(system, 3)))
@@ -258,7 +266,7 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, Handle*, u64, u32, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    u32 param_1 = 0;
    const u32 retval = func(system, &param_1, Param(system, 1), static_cast<u32>(Param(system, 2)),
                            static_cast<u32>(Param(system, 3)))
@@ -269,14 +277,14 @@ void SvcWrap(Core::System& system) {
 }
 template <ResultCode func(Core::System&, u64, u32, s32, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
                            static_cast<s32>(Param(system, 2)), static_cast<s64>(Param(system, 3)))
                           .raw);
 }
 template <ResultCode func(Core::System&, u64, u32, s32, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system, Param(system, 0), static_cast<u32>(Param(system, 1)),
                            static_cast<s32>(Param(system, 2)), static_cast<s32>(Param(system, 3)))
                           .raw);
@@ -286,7 +294,7 @@ void SvcWrap(Core::System& system) {
 // Function wrappers that return type u32
 template <u32 func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system));
 }
@@ -294,7 +302,7 @@ void SvcWrap(Core::System& system) {
 // Function wrappers that return type u64
 template <u64 func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    FuncReturn(system, func(system));
 }
@@ -302,44 +310,110 @@ void SvcWrap(Core::System& system) {
 /// Function wrappers that return type void
 template <void func(Core::System&)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system);
 }
 template <void func(Core::System&, u32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system, static_cast<u32>(Param(system, 0)));
 }
 template <void func(Core::System&, u32, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2),
         Param(system, 3));
 }
 template <void func(Core::System&, s64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system, static_cast<s64>(Param(system, 0)));
 }
 template <void func(Core::System&, u64, s32)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system, Param(system, 0), static_cast<s32>(Param(system, 1)));
 }
 template <void func(Core::System&, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system, Param(system, 0), Param(system, 1));
 }
 template <void func(Core::System&, u64, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system, Param(system, 0), Param(system, 1), Param(system, 2));
 }
 template <void func(Core::System&, u32, u64, u64)>
-void SvcWrap(Core::System& system) {
+void SvcWrap64(Core::System& system) {
    func(system, static_cast<u32>(Param(system, 0)), Param(system, 1), Param(system, 2));
 }
+// Used by QueryMemory32
+template <ResultCode func(Core::System&, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn32(system,
+                 func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2)).raw);
+}
+// Used by GetInfo32
+template <ResultCode func(Core::System&, u32*, u32*, u32, u32, u32, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 0), Param32(system, 1),
+                            Param32(system, 2), Param32(system, 3))
+                           .raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+// Used by GetThreadPriority32, ConnectToNamedPort32
+template <ResultCode func(Core::System&, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, &param_1, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
+// Used by GetThreadId32
+template <ResultCode func(Core::System&, u32*, u32*, u32)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    u32 param_2 = 0;
+    const u32 retval = func(system, &param_1, &param_2, Param32(system, 1)).raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    system.CurrentArmInterface().SetReg(2, param_2);
+    FuncReturn(system, retval);
+}
+// Used by SignalProcessWideKey32
+template <void func(Core::System&, u32, s32)>
+void SvcWrap32(Core::System& system) {
+    func(system, static_cast<u32>(Param(system, 0)), static_cast<s32>(Param(system, 1)));
+}
+// Used by SendSyncRequest32
+template <ResultCode func(Core::System&, u32)>
+void SvcWrap32(Core::System& system) {
+    FuncReturn(system, func(system, static_cast<u32>(Param(system, 0))).raw);
+}
+// Used by WaitSynchronization32
+template <ResultCode func(Core::System&, u32, u32, s32, u32, Handle*)>
+void SvcWrap32(Core::System& system) {
+    u32 param_1 = 0;
+    const u32 retval = func(system, Param32(system, 0), Param32(system, 1), Param32(system, 2),
+                            Param32(system, 3), &param_1)
+                           .raw;
+    system.CurrentArmInterface().SetReg(1, param_1);
+    FuncReturn(system, retval);
+}
 } // namespace Kernel
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index bf850e0b2..83e956036 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -133,15 +133,16 @@ void Thread::CancelWait() {
    ResumeFromWait();
 }
-/**
+static void ResetThreadContext32(Core::ARM_Interface::ThreadContext32& context, u32 stack_top,
- * Resets a thread context, making it ready to be scheduled and run by the CPU
+                                 u32 entry_point, u32 arg) {
- * @param context Thread context to reset
+    context = {};
- * @param stack_top Address of the top of the stack
+    context.cpu_registers[0] = arg;
- * @param entry_point Address of entry point for execution
+    context.cpu_registers[15] = entry_point;
- * @param arg User argument for thread
+    context.cpu_registers[13] = stack_top;
- */
+}
-static void ResetThreadContext(Core::ARM_Interface::ThreadContext& context, VAddr stack_top,
-                               VAddr entry_point, u64 arg) {
+static void ResetThreadContext64(Core::ARM_Interface::ThreadContext64& context, VAddr stack_top,
+                                 VAddr entry_point, u64 arg) {
    context = {};
    context.cpu_registers[0] = arg;
    context.pc = entry_point;
@@ -198,9 +199,9 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(KernelCore& kernel, std::strin
    thread->owner_process->RegisterThread(thread.get());
-    // TODO(peachum): move to ScheduleThread() when scheduler is added so selected core is used
+    ResetThreadContext32(thread->context_32, static_cast<u32>(stack_top),
-    // to initialize the context
+                         static_cast<u32>(entry_point), static_cast<u32>(arg));
-    ResetThreadContext(thread->context, stack_top, entry_point, arg);
+    ResetThreadContext64(thread->context_64, stack_top, entry_point, arg);
    return MakeResult<std::shared_ptr<Thread>>(std::move(thread));
 }
@@ -213,11 +214,13 @@ void Thread::SetPriority(u32 priority) {
 }
 void Thread::SetWaitSynchronizationResult(ResultCode result) {
-    context.cpu_registers[0] = result.raw;
+    context_32.cpu_registers[0] = result.raw;
+    context_64.cpu_registers[0] = result.raw;
 }
 void Thread::SetWaitSynchronizationOutput(s32 output) {
-    context.cpu_registers[1] = output;
+    context_32.cpu_registers[1] = output;
+    context_64.cpu_registers[1] = output;
 }
 s32 Thread::GetSynchronizationObjectIndex(std::shared_ptr<SynchronizationObject> object) const {
diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h
index 129e7858a..23fdef8a4 100644
--- a/src/core/hle/kernel/thread.h
+++ b/src/core/hle/kernel/thread.h
@@ -102,7 +102,8 @@ public:
    using MutexWaitingThreads = std::vector<std::shared_ptr<Thread>>;
-    using ThreadContext = Core::ARM_Interface::ThreadContext;
+    using ThreadContext32 = Core::ARM_Interface::ThreadContext32;
+    using ThreadContext64 = Core::ARM_Interface::ThreadContext64;
    using ThreadSynchronizationObjects = std::vector<std::shared_ptr<SynchronizationObject>>;
@@ -273,12 +274,20 @@ public:
        return status == ThreadStatus::WaitSynch;
    }
-    ThreadContext& GetContext() {
+    ThreadContext32& GetContext32() {
-        return context;
+        return context_32;
    }
-    const ThreadContext& GetContext() const {
+    const ThreadContext32& GetContext32() const {
-        return context;
+        return context_32;
+    }
+    ThreadContext64& GetContext64() {
+        return context_64;
+    }
+    const ThreadContext64& GetContext64() const {
+        return context_64;
    }
    ThreadStatus GetStatus() const {
@@ -466,7 +475,8 @@ private:
    void AdjustSchedulingOnPriority(u32 old_priority);
    void AdjustSchedulingOnAffinity(u64 old_affinity_mask, s32 old_core);
-    Core::ARM_Interface::ThreadContext context{};
+    ThreadContext32 context_32{};
+    ThreadContext64 context_64{};
    u64 thread_id = 0;
diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp
index cc978713b..d1bf13c89 100644
--- a/src/core/hle/service/am/am.cpp
+++ b/src/core/hle/service/am/am.cpp
@@ -607,7 +607,7 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system,
        {40, nullptr, "GetCradleFwVersion"},
        {50, nullptr, "IsVrModeEnabled"},
        {51, nullptr, "SetVrModeEnabled"},
-        {52, nullptr, "SwitchLcdBacklight"},
+        {52, &ICommonStateGetter::SetLcdBacklighOffEnabled, "SetLcdBacklighOffEnabled"},
        {53, nullptr, "BeginVrModeEx"},
        {54, nullptr, "EndVrModeEx"},
        {55, nullptr, "IsInControllerFirmwareUpdateSection"},
@@ -636,7 +636,6 @@ void ICommonStateGetter::GetBootMode(Kernel::HLERequestContext& ctx) {
    IPC::ResponseBuilder rb{ctx, 3};
    rb.Push(RESULT_SUCCESS);
    rb.Push<u8>(static_cast<u8>(Service::PM::SystemBootMode::Normal)); // Normal boot mode
 }
@@ -660,6 +659,7 @@ void ICommonStateGetter::ReceiveMessage(Kernel::HLERequestContext& ctx) {
        rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
        return;
    }
    rb.Push(RESULT_SUCCESS);
    rb.PushEnum<AppletMessageQueue::AppletMessage>(message);
 }
@@ -672,6 +672,17 @@ void ICommonStateGetter::GetCurrentFocusState(Kernel::HLERequestContext& ctx) {
    rb.Push(static_cast<u8>(FocusState::InFocus));
 }
+void ICommonStateGetter::SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto is_lcd_backlight_off_enabled = rp.Pop<bool>();
+    LOG_WARNING(Service_AM, "(STUBBED) called. is_lcd_backlight_off_enabled={}",
+                is_lcd_backlight_off_enabled);
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
 void ICommonStateGetter::GetDefaultDisplayResolutionChangeEvent(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service_AM, "called");
diff --git a/src/core/hle/service/am/am.h b/src/core/hle/service/am/am.h
index 0b9a4332d..0843de781 100644
--- a/src/core/hle/service/am/am.h
+++ b/src/core/hle/service/am/am.h
@@ -182,6 +182,7 @@ private:
    void GetOperationMode(Kernel::HLERequestContext& ctx);
    void GetPerformanceMode(Kernel::HLERequestContext& ctx);
    void GetBootMode(Kernel::HLERequestContext& ctx);
+    void SetLcdBacklighOffEnabled(Kernel::HLERequestContext& ctx);
    void GetDefaultDisplayResolution(Kernel::HLERequestContext& ctx);
    void SetCpuBoostMode(Kernel::HLERequestContext& ctx);
diff --git a/src/core/loader/deconstructed_rom_directory.cpp b/src/core/loader/deconstructed_rom_directory.cpp
index d19c3623c..53559e8b1 100644
--- a/src/core/loader/deconstructed_rom_directory.cpp
+++ b/src/core/loader/deconstructed_rom_directory.cpp
@@ -129,12 +129,6 @@ AppLoader_DeconstructedRomDirectory::LoadResult AppLoader_DeconstructedRomDirect
    }
    metadata.Print();
-    const FileSys::ProgramAddressSpaceType arch_bits{metadata.GetAddressSpaceType()};
-    if (arch_bits == FileSys::ProgramAddressSpaceType::Is32Bit ||
-        arch_bits == FileSys::ProgramAddressSpaceType::Is32BitNoMap) {
-        return {ResultStatus::Error32BitISA, {}};
-    }
    if (process.LoadFromMetadata(metadata).IsError()) {
        return {ResultStatus::ErrorUnableToParseKernelMetadata, {}};
    }
diff --git a/src/core/reporter.cpp b/src/core/reporter.cpp
index f95eee3b1..85ac81ef7 100644
--- a/src/core/reporter.cpp
+++ b/src/core/reporter.cpp
@@ -111,7 +111,7 @@ json GetProcessorStateDataAuto(Core::System& system) {
    const auto& vm_manager{process->VMManager()};
    auto& arm{system.CurrentArmInterface()};
-    Core::ARM_Interface::ThreadContext context{};
+    Core::ARM_Interface::ThreadContext64 context{};
    arm.SaveContext(context);
    return GetProcessorStateData(process->Is64BitProcess() ? "AArch64" : "AArch32",
diff --git a/src/core/settings.cpp b/src/core/settings.cpp
index d1fc94060..7c0303684 100644
--- a/src/core/settings.cpp
+++ b/src/core/settings.cpp
@@ -94,6 +94,7 @@ void LogSettings() {
    LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
    LogSetting("Renderer_UseAsynchronousGpuEmulation",
               Settings::values.use_asynchronous_gpu_emulation);
+    LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
    LogSetting("Audio_OutputEngine", Settings::values.sink_id);
    LogSetting("Audio_EnableAudioStretching", Settings::values.enable_audio_stretching);
    LogSetting("Audio_OutputDevice", Settings::values.audio_device_id);
diff --git a/src/core/settings.h b/src/core/settings.h
index f837d3fbc..15b691342 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -435,6 +435,7 @@ struct Values {
    bool use_disk_shader_cache;
    bool use_accurate_gpu_emulation;
    bool use_asynchronous_gpu_emulation;
+    bool use_vsync;
    bool force_30fps_mode;
    float bg_red;
diff --git a/src/core/telemetry_session.cpp b/src/core/telemetry_session.cpp
index 0e72d31cd..0f3685d1c 100644
--- a/src/core/telemetry_session.cpp
+++ b/src/core/telemetry_session.cpp
@@ -188,6 +188,7 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
             Settings::values.use_accurate_gpu_emulation);
    AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
             Settings::values.use_asynchronous_gpu_emulation);
+    AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync);
    AddField(field_type, "System_UseDockedMode", Settings::values.use_docked_mode);
 }
diff --git a/src/input_common/udp/client.cpp b/src/input_common/udp/client.cpp
index 2228571a6..e82ae7ef1 100644
--- a/src/input_common/udp/client.cpp
+++ b/src/input_common/udp/client.cpp
@@ -32,8 +32,16 @@ public:
                    SocketCallback callback)
        : callback(std::move(callback)), timer(io_service),
          socket(io_service, udp::endpoint(udp::v4(), 0)), client_id(client_id),
-          pad_index(pad_index),
+          pad_index(pad_index) {
-          send_endpoint(udp::endpoint(boost::asio::ip::make_address_v4(host), port)) {}
+        boost::system::error_code ec{};
+        auto ipv4 = boost::asio::ip::make_address_v4(host, ec);
+        if (ec.failed()) {
+            LOG_ERROR(Input, "Invalid IPv4 address \"{}\" provided to socket", host);
+            ipv4 = boost::asio::ip::address_v4{};
+        }
+        send_endpoint = {udp::endpoint(ipv4, port)};
+    }
    void Stop() {
        io_service.stop();
@@ -85,17 +93,18 @@ private:
    }
    void HandleSend(const boost::system::error_code& error) {
+        boost::system::error_code _ignored{};
        // Send a request for getting port info for the pad
        Request::PortInfo port_info{1, {pad_index, 0, 0, 0}};
        const auto port_message = Request::Create(port_info, client_id);
        std::memcpy(&send_buffer1, &port_message, PORT_INFO_SIZE);
-        socket.send_to(boost::asio::buffer(send_buffer1), send_endpoint);
+        socket.send_to(boost::asio::buffer(send_buffer1), send_endpoint, {}, _ignored);
        // Send a request for getting pad data for the pad
        Request::PadData pad_data{Request::PadData::Flags::Id, pad_index, EMPTY_MAC_ADDRESS};
        const auto pad_message = Request::Create(pad_data, client_id);
        std::memcpy(send_buffer2.data(), &pad_message, PAD_DATA_SIZE);
-        socket.send_to(boost::asio::buffer(send_buffer2), send_endpoint);
+        socket.send_to(boost::asio::buffer(send_buffer2), send_endpoint, {}, _ignored);
        StartSend(timer.expiry());
    }
diff --git a/src/input_common/udp/protocol.cpp b/src/input_common/udp/protocol.cpp
index a982ac49d..5e50bd612 100644
--- a/src/input_common/udp/protocol.cpp
+++ b/src/input_common/udp/protocol.cpp
@@ -31,7 +31,6 @@ namespace Response {
 */
 std::optional<Type> Validate(u8* data, std::size_t size) {
    if (size < sizeof(Header)) {
-        LOG_DEBUG(Input, "Invalid UDP packet received");
        return std::nullopt;
    }
    Header header{};
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index 2cdf1aa7f..b1088af3d 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -5,7 +5,7 @@
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "core/core.h"
-#include "core/frontend/scope_acquire_window_context.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/gpu.h"
 #include "video_core/gpu_thread.h"
@@ -27,7 +27,7 @@ static void RunThread(VideoCore::RendererBase& renderer, Tegra::DmaPusher& dma_p
        return;
    }
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{renderer.GetRenderWindow()};
+    Core::Frontend::ScopeAcquireContext acquire_context{renderer.GetRenderWindow()};
    CommandDataContainer next;
    while (state.is_running) {
diff --git a/src/video_core/renderer_base.h b/src/video_core/renderer_base.h
index af1bebc4f..5ec99a126 100644
--- a/src/video_core/renderer_base.h
+++ b/src/video_core/renderer_base.h
@@ -35,15 +35,19 @@ public:
    explicit RendererBase(Core::Frontend::EmuWindow& window);
    virtual ~RendererBase();
-    /// Swap buffers (render frame)
-    virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
    /// Initialize the renderer
    virtual bool Init() = 0;
    /// Shutdown the renderer
    virtual void ShutDown() = 0;
+    /// Finalize rendering the guest frame and draw into the presentation texture
+    virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
+    /// Draws the latest frame to the window waiting timeout_ms for a frame to arrive (Renderer
+    /// specific implementation)
+    virtual void TryPresent(int timeout_ms) = 0;
    // Getter/setter functions:
    // ------------------------
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index e1965fb21..3fcd319fd 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -36,6 +36,7 @@ namespace OpenGL {
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+using Tegra::Engines::ShaderType;
 using VideoCore::Surface::PixelFormat;
 using VideoCore::Surface::SurfaceTarget;
 using VideoCore::Surface::SurfaceType;
@@ -56,8 +57,7 @@ namespace {
 template <typename Engine, typename Entry>
 Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
-                                               Tegra::Engines::ShaderType shader_type,
+                                               ShaderType shader_type, std::size_t index = 0) {
-                                               std::size_t index = 0) {
    if (entry.IsBindless()) {
        const Tegra::Texture::TextureHandle tex_handle =
            engine.AccessConstBuffer32(shader_type, entry.GetBuffer(), entry.GetOffset());
@@ -910,15 +910,10 @@ void RasterizerOpenGL::SetupDrawTextures(std::size_t stage_index, const Shader&
    const auto& maxwell3d = system.GPU().Maxwell3D();
    u32 binding = device.GetBaseBindings(stage_index).sampler;
    for (const auto& entry : shader->GetShaderEntries().samplers) {
-        const auto shader_type = static_cast<Tegra::Engines::ShaderType>(stage_index);
+        const auto shader_type = static_cast<ShaderType>(stage_index);
-        if (!entry.IsIndexed()) {
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
-            const auto texture = GetTextureInfo(maxwell3d, entry, shader_type);
+            const auto texture = GetTextureInfo(maxwell3d, entry, shader_type, i);
            SetupTexture(binding++, texture, entry);
-        } else {
-            for (std::size_t i = 0; i < entry.Size(); ++i) {
-                const auto texture = GetTextureInfo(maxwell3d, entry, shader_type, i);
-                SetupTexture(binding++, texture, entry);
-            }
        }
    }
 }
@@ -928,16 +923,9 @@ void RasterizerOpenGL::SetupComputeTextures(const Shader& kernel) {
    const auto& compute = system.GPU().KeplerCompute();
    u32 binding = 0;
    for (const auto& entry : kernel->GetShaderEntries().samplers) {
-        if (!entry.IsIndexed()) {
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
-            const auto texture =
+            const auto texture = GetTextureInfo(compute, entry, ShaderType::Compute, i);
-                GetTextureInfo(compute, entry, Tegra::Engines::ShaderType::Compute);
            SetupTexture(binding++, texture, entry);
-        } else {
-            for (std::size_t i = 0; i < entry.Size(); ++i) {
-                const auto texture =
-                    GetTextureInfo(compute, entry, Tegra::Engines::ShaderType::Compute, i);
-                SetupTexture(binding++, texture, entry);
-            }
        }
    }
 }
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp
index f0ddfb276..c0aee770f 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp
@@ -15,6 +15,24 @@ MICROPROFILE_DEFINE(OpenGL_ResourceDeletion, "OpenGL", "Resource Deletion", MP_R
 namespace OpenGL {
+void OGLRenderbuffer::Create() {
+    if (handle != 0)
+        return;
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
+    glGenRenderbuffers(1, &handle);
+}
+void OGLRenderbuffer::Release() {
+    if (handle == 0)
+        return;
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
+    glDeleteRenderbuffers(1, &handle);
+    OpenGLState::GetCurState().ResetRenderbuffer(handle).Apply();
+    handle = 0;
+}
 void OGLTexture::Create(GLenum target) {
    if (handle != 0)
        return;
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h
index 514d1d165..995a4e45e 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.h
+++ b/src/video_core/renderer_opengl/gl_resource_manager.h
@@ -11,6 +11,31 @@
 namespace OpenGL {
+class OGLRenderbuffer : private NonCopyable {
+public:
+    OGLRenderbuffer() = default;
+    OGLRenderbuffer(OGLRenderbuffer&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+    ~OGLRenderbuffer() {
+        Release();
+    }
+    OGLRenderbuffer& operator=(OGLRenderbuffer&& o) noexcept {
+        Release();
+        handle = std::exchange(o.handle, 0);
+        return *this;
+    }
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create();
+    /// Deletes the internal OpenGL resource
+    void Release();
+    GLuint handle = 0;
+};
 class OGLTexture : private NonCopyable {
 public:
    OGLTexture() = default;
diff --git a/src/video_core/renderer_opengl/gl_state.cpp b/src/video_core/renderer_opengl/gl_state.cpp
index ab1f7983c..7d3bc1a1f 100644
--- a/src/video_core/renderer_opengl/gl_state.cpp
+++ b/src/video_core/renderer_opengl/gl_state.cpp
@@ -423,6 +423,13 @@ void OpenGLState::ApplyClipControl() {
    }
 }
+void OpenGLState::ApplyRenderBuffer() {
+    if (cur_state.renderbuffer != renderbuffer) {
+        cur_state.renderbuffer = renderbuffer;
+        glBindRenderbuffer(GL_RENDERBUFFER, renderbuffer);
+    }
+}
 void OpenGLState::ApplyTextures() {
    const std::size_t size = std::size(textures);
    for (std::size_t i = 0; i < size; ++i) {
@@ -478,6 +485,7 @@ void OpenGLState::Apply() {
    ApplyPolygonOffset();
    ApplyAlphaTest();
    ApplyClipControl();
+    ApplyRenderBuffer();
 }
 void OpenGLState::EmulateViewportWithScissor() {
@@ -551,4 +559,11 @@ OpenGLState& OpenGLState::ResetFramebuffer(GLuint handle) {
    return *this;
 }
+OpenGLState& OpenGLState::ResetRenderbuffer(GLuint handle) {
+    if (renderbuffer == handle) {
+        renderbuffer = 0;
+    }
+    return *this;
+}
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_state.h b/src/video_core/renderer_opengl/gl_state.h
index 4953eeda2..bce662f2c 100644
--- a/src/video_core/renderer_opengl/gl_state.h
+++ b/src/video_core/renderer_opengl/gl_state.h
@@ -158,6 +158,8 @@ public:
        GLenum depth_mode = GL_NEGATIVE_ONE_TO_ONE;
    } clip_control;
+    GLuint renderbuffer{}; // GL_RENDERBUFFER_BINDING
    OpenGLState();
    /// Get the currently active OpenGL state
@@ -196,6 +198,7 @@ public:
    void ApplyPolygonOffset();
    void ApplyAlphaTest();
    void ApplyClipControl();
+    void ApplyRenderBuffer();
    /// Resets any references to the given resource
    OpenGLState& UnbindTexture(GLuint handle);
@@ -204,6 +207,7 @@ public:
    OpenGLState& ResetPipeline(GLuint handle);
    OpenGLState& ResetVertexArray(GLuint handle);
    OpenGLState& ResetFramebuffer(GLuint handle);
+    OpenGLState& ResetRenderbuffer(GLuint handle);
    /// Viewport does not affects glClearBuffer so emulate viewport using scissor test
    void EmulateViewportWithScissor();
diff --git a/src/video_core/renderer_opengl/maxwell_to_gl.h b/src/video_core/renderer_opengl/maxwell_to_gl.h
index 7ed505628..d3dea3659 100644
--- a/src/video_core/renderer_opengl/maxwell_to_gl.h
+++ b/src/video_core/renderer_opengl/maxwell_to_gl.h
@@ -92,8 +92,32 @@ inline GLenum VertexType(Maxwell::VertexAttribute attrib) {
        }
    case Maxwell::VertexAttribute::Type::UnsignedScaled:
        switch (attrib.size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
        case Maxwell::VertexAttribute::Size::Size_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
            return GL_UNSIGNED_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return GL_UNSIGNED_SHORT;
+        default:
+            LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
+            return {};
+        }
+    case Maxwell::VertexAttribute::Type::SignedScaled:
+        switch (attrib.size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return GL_BYTE;
+        case Maxwell::VertexAttribute::Size::Size_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return GL_SHORT;
        default:
            LOG_ERROR(Render_OpenGL, "Unimplemented vertex size={}", attrib.SizeString());
            return {};
diff --git a/src/video_core/renderer_opengl/renderer_opengl.cpp b/src/video_core/renderer_opengl/renderer_opengl.cpp
index bba16afaf..a4340b502 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.cpp
+++ b/src/video_core/renderer_opengl/renderer_opengl.cpp
@@ -9,11 +9,11 @@
 #include <glad/glad.h>
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/microprofile.h"
 #include "common/telemetry.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
-#include "core/frontend/scope_acquire_window_context.h"
 #include "core/memory.h"
 #include "core/perf_stats.h"
 #include "core/settings.h"
@@ -24,6 +24,144 @@
 namespace OpenGL {
+// If the size of this is too small, it ends up creating a soft cap on FPS as the renderer will have
+// to wait on available presentation frames.
+constexpr std::size_t SWAP_CHAIN_SIZE = 3;
+struct Frame {
+    u32 width{};                      /// Width of the frame (to detect resize)
+    u32 height{};                     /// Height of the frame
+    bool color_reloaded{};            /// Texture attachment was recreated (ie: resized)
+    OpenGL::OGLRenderbuffer color{};  /// Buffer shared between the render/present FBO
+    OpenGL::OGLFramebuffer render{};  /// FBO created on the render thread
+    OpenGL::OGLFramebuffer present{}; /// FBO created on the present thread
+    GLsync render_fence{};            /// Fence created on the render thread
+    GLsync present_fence{};           /// Fence created on the presentation thread
+    bool is_srgb{};                   /// Framebuffer is sRGB or RGB
+};
+/**
+ * For smooth Vsync rendering, we want to always present the latest frame that the core generates,
+ * but also make sure that rendering happens at the pace that the frontend dictates. This is a
+ * helper class that the renderer uses to sync frames between the render thread and the presentation
+ * thread
+ */
+class FrameMailbox {
+public:
+    std::mutex swap_chain_lock;
+    std::condition_variable present_cv;
+    std::array<Frame, SWAP_CHAIN_SIZE> swap_chain{};
+    std::queue<Frame*> free_queue;
+    std::deque<Frame*> present_queue;
+    Frame* previous_frame{};
+    FrameMailbox() {
+        for (auto& frame : swap_chain) {
+            free_queue.push(&frame);
+        }
+    }
+    ~FrameMailbox() {
+        // lock the mutex and clear out the present and free_queues and notify any people who are
+        // blocked to prevent deadlock on shutdown
+        std::scoped_lock lock{swap_chain_lock};
+        std::queue<Frame*>().swap(free_queue);
+        present_queue.clear();
+        present_cv.notify_all();
+    }
+    void ReloadPresentFrame(Frame* frame, u32 height, u32 width) {
+        frame->present.Release();
+        frame->present.Create();
+        GLint previous_draw_fbo{};
+        glGetIntegerv(GL_DRAW_FRAMEBUFFER_BINDING, &previous_draw_fbo);
+        glBindFramebuffer(GL_FRAMEBUFFER, frame->present.handle);
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate present FBO!");
+        }
+        glBindFramebuffer(GL_DRAW_FRAMEBUFFER, previous_draw_fbo);
+        frame->color_reloaded = false;
+    }
+    void ReloadRenderFrame(Frame* frame, u32 width, u32 height) {
+        OpenGLState prev_state = OpenGLState::GetCurState();
+        OpenGLState state = OpenGLState::GetCurState();
+        // Recreate the color texture attachment
+        frame->color.Release();
+        frame->color.Create();
+        state.renderbuffer = frame->color.handle;
+        state.Apply();
+        glRenderbufferStorage(GL_RENDERBUFFER, frame->is_srgb ? GL_SRGB8 : GL_RGB8, width, height);
+        // Recreate the FBO for the render target
+        frame->render.Release();
+        frame->render.Create();
+        state.draw.read_framebuffer = frame->render.handle;
+        state.draw.draw_framebuffer = frame->render.handle;
+        state.Apply();
+        glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER,
+                                  frame->color.handle);
+        if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE) {
+            LOG_CRITICAL(Render_OpenGL, "Failed to recreate render FBO!");
+        }
+        prev_state.Apply();
+        frame->width = width;
+        frame->height = height;
+        frame->color_reloaded = true;
+    }
+    Frame* GetRenderFrame() {
+        std::unique_lock lock{swap_chain_lock};
+        // If theres no free frames, we will reuse the oldest render frame
+        if (free_queue.empty()) {
+            auto frame = present_queue.back();
+            present_queue.pop_back();
+            return frame;
+        }
+        Frame* frame = free_queue.front();
+        free_queue.pop();
+        return frame;
+    }
+    void ReleaseRenderFrame(Frame* frame) {
+        std::unique_lock lock{swap_chain_lock};
+        present_queue.push_front(frame);
+        present_cv.notify_one();
+    }
+    Frame* TryGetPresentFrame(int timeout_ms) {
+        std::unique_lock lock{swap_chain_lock};
+        // wait for new entries in the present_queue
+        present_cv.wait_for(lock, std::chrono::milliseconds(timeout_ms),
+                            [&] { return !present_queue.empty(); });
+        if (present_queue.empty()) {
+            // timed out waiting for a frame to draw so return the previous frame
+            return previous_frame;
+        }
+        // free the previous frame and add it back to the free queue
+        if (previous_frame) {
+            free_queue.push(previous_frame);
+        }
+        // the newest entries are pushed to the front of the queue
+        Frame* frame = present_queue.front();
+        present_queue.pop_front();
+        // remove all old entries from the present queue and move them back to the free_queue
+        for (auto f : present_queue) {
+            free_queue.push(f);
+        }
+        present_queue.clear();
+        previous_frame = frame;
+        return frame;
+    }
+};
 namespace {
 constexpr char vertex_shader[] = R"(
@@ -158,21 +296,91 @@ void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severit
 } // Anonymous namespace
 RendererOpenGL::RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system)
-    : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system} {}
+    : VideoCore::RendererBase{emu_window}, emu_window{emu_window}, system{system},
+      frame_mailbox{std::make_unique<FrameMailbox>()} {}
 RendererOpenGL::~RendererOpenGL() = default;
+MICROPROFILE_DEFINE(OpenGL_RenderFrame, "OpenGL", "Render Frame", MP_RGB(128, 128, 64));
+MICROPROFILE_DEFINE(OpenGL_WaitPresent, "OpenGL", "Wait For Present", MP_RGB(128, 128, 128));
 void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
+    render_window.PollEvents();
+    if (!framebuffer) {
+        return;
+    }
    // Maintain the rasterizer's state as a priority
    OpenGLState prev_state = OpenGLState::GetCurState();
    state.AllDirty();
    state.Apply();
+    PrepareRendertarget(framebuffer);
+    RenderScreenshot();
+    Frame* frame;
+    {
+        MICROPROFILE_SCOPE(OpenGL_WaitPresent);
+        frame = frame_mailbox->GetRenderFrame();
+        // Clean up sync objects before drawing
+        // INTEL driver workaround. We can't delete the previous render sync object until we are
+        // sure that the presentation is done
+        if (frame->present_fence) {
+            glClientWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
+        }
+        // delete the draw fence if the frame wasn't presented
+        if (frame->render_fence) {
+            glDeleteSync(frame->render_fence);
+            frame->render_fence = 0;
+        }
+        // wait for the presentation to be done
+        if (frame->present_fence) {
+            glWaitSync(frame->present_fence, 0, GL_TIMEOUT_IGNORED);
+            glDeleteSync(frame->present_fence);
+            frame->present_fence = 0;
+        }
+    }
+    {
+        MICROPROFILE_SCOPE(OpenGL_RenderFrame);
+        const auto& layout = render_window.GetFramebufferLayout();
+        // Recreate the frame if the size of the window has changed
+        if (layout.width != frame->width || layout.height != frame->height ||
+            screen_info.display_srgb != frame->is_srgb) {
+            LOG_DEBUG(Render_OpenGL, "Reloading render frame");
+            frame->is_srgb = screen_info.display_srgb;
+            frame_mailbox->ReloadRenderFrame(frame, layout.width, layout.height);
+        }
+        state.draw.draw_framebuffer = frame->render.handle;
+        state.Apply();
+        DrawScreen(layout);
+        // Create a fence for the frontend to wait on and swap this frame to OffTex
+        frame->render_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+        glFlush();
+        frame_mailbox->ReleaseRenderFrame(frame);
+        m_current_frame++;
+        rasterizer->TickFrame();
+    }
+    // Restore the rasterizer state
+    prev_state.AllDirty();
+    prev_state.Apply();
+}
+void RendererOpenGL::PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer) {
    if (framebuffer) {
        // If framebuffer is provided, reload it from memory to a texture
        if (screen_info.texture.width != static_cast<GLsizei>(framebuffer->width) ||
            screen_info.texture.height != static_cast<GLsizei>(framebuffer->height) ||
-            screen_info.texture.pixel_format != framebuffer->pixel_format) {
+            screen_info.texture.pixel_format != framebuffer->pixel_format ||
+            gl_framebuffer_data.empty()) {
            // Reallocate texture if the framebuffer size has changed.
            // This is expected to not happen very often and hence should not be a
            // performance problem.
@@ -181,22 +389,7 @@ void RendererOpenGL::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
        // Load the framebuffer from memory, draw it to the screen, and swap buffers
        LoadFBToScreenInfo(*framebuffer);
-        if (renderer_settings.screenshot_requested)
-            CaptureScreenshot();
-        DrawScreen(render_window.GetFramebufferLayout());
-        rasterizer->TickFrame();
-        render_window.SwapBuffers();
    }
-    render_window.PollEvents();
-    // Restore the rasterizer state
-    prev_state.AllDirty();
-    prev_state.Apply();
 }
 void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer) {
@@ -418,13 +611,48 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
    DrawScreenTriangles(screen_info, static_cast<float>(screen.left),
                        static_cast<float>(screen.top), static_cast<float>(screen.GetWidth()),
                        static_cast<float>(screen.GetHeight()));
+}
-    m_current_frame++;
+void RendererOpenGL::TryPresent(int timeout_ms) {
+    const auto& layout = render_window.GetFramebufferLayout();
+    auto frame = frame_mailbox->TryGetPresentFrame(timeout_ms);
+    if (!frame) {
+        LOG_DEBUG(Render_OpenGL, "TryGetPresentFrame returned no frame to present");
+        return;
+    }
+    // Clearing before a full overwrite of a fbo can signal to drivers that they can avoid a
+    // readback since we won't be doing any blending
+    glClear(GL_COLOR_BUFFER_BIT);
+    // Recreate the presentation FBO if the color attachment was changed
+    if (frame->color_reloaded) {
+        LOG_DEBUG(Render_OpenGL, "Reloading present frame");
+        frame_mailbox->ReloadPresentFrame(frame, layout.width, layout.height);
+    }
+    glWaitSync(frame->render_fence, 0, GL_TIMEOUT_IGNORED);
+    // INTEL workaround.
+    // Normally we could just delete the draw fence here, but due to driver bugs, we can just delete
+    // it on the emulation thread without too much penalty
+    // glDeleteSync(frame.render_sync);
+    // frame.render_sync = 0;
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, frame->present.handle);
+    glBlitFramebuffer(0, 0, frame->width, frame->height, 0, 0, layout.width, layout.height,
+                      GL_COLOR_BUFFER_BIT, GL_LINEAR);
+    // Insert fence for the main thread to block on
+    frame->present_fence = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+    glFlush();
+    glBindFramebuffer(GL_READ_FRAMEBUFFER, 0);
 }
-void RendererOpenGL::UpdateFramerate() {}
+void RendererOpenGL::RenderScreenshot() {
+    if (!renderer_settings.screenshot_requested) {
+        return;
+    }
-void RendererOpenGL::CaptureScreenshot() {
    // Draw the current frame to the screenshot framebuffer
    screenshot_framebuffer.Create();
    GLuint old_read_fb = state.draw.read_framebuffer;
@@ -459,8 +687,6 @@ void RendererOpenGL::CaptureScreenshot() {
 }
 bool RendererOpenGL::Init() {
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{render_window};
    if (GLAD_GL_KHR_debug) {
        glEnable(GL_DEBUG_OUTPUT);
        glDebugMessageCallback(DebugHandler, nullptr);
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h
index b56328a7f..d45e69cbc 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.h
+++ b/src/video_core/renderer_opengl/renderer_opengl.h
@@ -44,19 +44,23 @@ struct ScreenInfo {
    TextureInfo texture;
 };
+struct PresentationTexture {
+    u32 width = 0;
+    u32 height = 0;
+    OGLTexture texture;
+};
+class FrameMailbox;
 class RendererOpenGL final : public VideoCore::RendererBase {
 public:
    explicit RendererOpenGL(Core::Frontend::EmuWindow& emu_window, Core::System& system);
    ~RendererOpenGL() override;
-    /// Swap buffers (render frame)
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    /// Initialize the renderer
    bool Init() override;
-    /// Shutdown the renderer
    void ShutDown() override;
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
+    void TryPresent(int timeout_ms) override;
 private:
    /// Initializes the OpenGL state and creates persistent objects.
@@ -74,10 +78,7 @@ private:
    void DrawScreenTriangles(const ScreenInfo& screen_info, float x, float y, float w, float h);
-    /// Updates the framerate.
+    void RenderScreenshot();
-    void UpdateFramerate();
-    void CaptureScreenshot();
    /// Loads framebuffer from emulated memory into the active OpenGL texture.
    void LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuffer);
@@ -87,6 +88,8 @@ private:
    void LoadColorToActiveGLTexture(u8 color_r, u8 color_g, u8 color_b, u8 color_a,
                                    const TextureInfo& texture);
+    void PrepareRendertarget(const Tegra::FramebufferConfig* framebuffer);
    Core::Frontend::EmuWindow& emu_window;
    Core::System& system;
@@ -107,6 +110,9 @@ private:
    /// Used for transforming the framebuffer orientation
    Tegra::FramebufferConfig::TransformFlags framebuffer_transform_flags;
    Common::Rectangle<int> framebuffer_crop_rect;
+    /// Frame presentation mailbox
+    std::unique_ptr<FrameMailbox> frame_mailbox;
 };
 } // namespace OpenGL
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index ef66dd141..aad0c895b 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -371,8 +371,22 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
        }
    case Maxwell::VertexAttribute::Type::UnsignedScaled:
        switch (size) {
+        case Maxwell::VertexAttribute::Size::Size_8:
+            return vk::Format::eR8Uscaled;
        case Maxwell::VertexAttribute::Size::Size_8_8:
            return vk::Format::eR8G8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8:
+            return vk::Format::eR8G8B8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_8_8_8_8:
+            return vk::Format::eR8G8B8A8Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16:
+            return vk::Format::eR16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16:
+            return vk::Format::eR16G16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16:
+            return vk::Format::eR16G16B16Uscaled;
+        case Maxwell::VertexAttribute::Size::Size_16_16_16_16:
+            return vk::Format::eR16G16B16A16Uscaled;
        default:
            break;
        }
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
index d5032b432..ddc62bc97 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp
@@ -106,8 +106,14 @@ RendererVulkan::~RendererVulkan() {
 }
 void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
+    render_window.PollEvents();
+    if (!framebuffer) {
+        return;
+    }
    const auto& layout = render_window.GetFramebufferLayout();
-    if (framebuffer && layout.width > 0 && layout.height > 0 && render_window.IsShown()) {
+    if (layout.width > 0 && layout.height > 0 && render_window.IsShown()) {
        const VAddr framebuffer_addr = framebuffer->address + framebuffer->offset;
        const bool use_accelerated =
            rasterizer->AccelerateDisplay(*framebuffer, framebuffer_addr, framebuffer->stride);
@@ -128,13 +134,16 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
            blit_screen->Recreate();
        }
-        render_window.SwapBuffers();
        rasterizer->TickFrame();
    }
    render_window.PollEvents();
 }
+void RendererVulkan::TryPresent(int /*timeout_ms*/) {
+    // TODO (bunnei): ImplementMe
+}
 bool RendererVulkan::Init() {
    PFN_vkGetInstanceProcAddr vkGetInstanceProcAddr{};
    render_window.RetrieveVulkanHandlers(&vkGetInstanceProcAddr, &instance, &surface);
@@ -262,4 +271,4 @@ void RendererVulkan::Report() const {
    telemetry_session.AddField(field, "GPU_Vulkan_Extensions", extensions);
 }
-} // namespace Vulkan
-\ No newline at end of file
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h
index a472c5dc9..f513397f0 100644
--- a/src/video_core/renderer_vulkan/renderer_vulkan.h
+++ b/src/video_core/renderer_vulkan/renderer_vulkan.h
@@ -36,14 +36,10 @@ public:
    explicit RendererVulkan(Core::Frontend::EmuWindow& window, Core::System& system);
    ~RendererVulkan() override;
-    /// Swap buffers (render frame)
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    /// Initialize the renderer
    bool Init() override;
-    /// Shutdown the renderer
    void ShutDown() override;
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
+    void TryPresent(int timeout_ms) override;
 private:
    std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(
diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
index 9d5b8de7a..60f57d83e 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
@@ -73,7 +73,7 @@ UniqueDescriptorUpdateTemplate VKComputePipeline::CreateDescriptorUpdateTemplate
    std::vector<vk::DescriptorUpdateTemplateEntry> template_entries;
    u32 binding = 0;
    u32 offset = 0;
-    FillDescriptorUpdateTemplateEntries(device, entries, binding, offset, template_entries);
+    FillDescriptorUpdateTemplateEntries(entries, binding, offset, template_entries);
    if (template_entries.empty()) {
        // If the shader doesn't use descriptor sets, skip template creation.
        return UniqueDescriptorUpdateTemplate{};
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
index b155dfb49..6a02403c1 100644
--- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
@@ -97,8 +97,7 @@ UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplat
    u32 offset = 0;
    for (const auto& stage : program) {
        if (stage) {
-            FillDescriptorUpdateTemplateEntries(device, stage->entries, binding, offset,
+            FillDescriptorUpdateTemplateEntries(stage->entries, binding, offset, template_entries);
-                                                template_entries);
        }
    }
    if (template_entries.empty()) {
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
index 7ddf7d3ee..696e4b291 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
@@ -36,6 +36,13 @@ using Tegra::Engines::ShaderType;
 namespace {
+// C++20's using enum
+constexpr auto eUniformBuffer = vk::DescriptorType::eUniformBuffer;
+constexpr auto eStorageBuffer = vk::DescriptorType::eStorageBuffer;
+constexpr auto eUniformTexelBuffer = vk::DescriptorType::eUniformTexelBuffer;
+constexpr auto eCombinedImageSampler = vk::DescriptorType::eCombinedImageSampler;
+constexpr auto eStorageImage = vk::DescriptorType::eStorageImage;
 constexpr VideoCommon::Shader::CompilerSettings compiler_settings{
    VideoCommon::Shader::CompileDepth::FullDecompile};
@@ -119,23 +126,32 @@ ShaderType GetShaderType(Maxwell::ShaderProgram program) {
    }
 }
+template <vk::DescriptorType descriptor_type, class Container>
+void AddBindings(std::vector<vk::DescriptorSetLayoutBinding>& bindings, u32& binding,
+                 vk::ShaderStageFlags stage_flags, const Container& container) {
+    const u32 num_entries = static_cast<u32>(std::size(container));
+    for (std::size_t i = 0; i < num_entries; ++i) {
+        u32 count = 1;
+        if constexpr (descriptor_type == eCombinedImageSampler) {
+            // Combined image samplers can be arrayed.
+            count = container[i].Size();
+        }
+        bindings.emplace_back(binding++, descriptor_type, count, stage_flags, nullptr);
+    }
+}
 u32 FillDescriptorLayout(const ShaderEntries& entries,
                         std::vector<vk::DescriptorSetLayoutBinding>& bindings,
                         Maxwell::ShaderProgram program_type, u32 base_binding) {
    const ShaderType stage = GetStageFromProgram(program_type);
-    const vk::ShaderStageFlags stage_flags = MaxwellToVK::ShaderStage(stage);
+    const vk::ShaderStageFlags flags = MaxwellToVK::ShaderStage(stage);
    u32 binding = base_binding;
-    const auto AddBindings = [&](vk::DescriptorType descriptor_type, std::size_t num_entries) {
+    AddBindings<eUniformBuffer>(bindings, binding, flags, entries.const_buffers);
-        for (std::size_t i = 0; i < num_entries; ++i) {
+    AddBindings<eStorageBuffer>(bindings, binding, flags, entries.global_buffers);
-            bindings.emplace_back(binding++, descriptor_type, 1, stage_flags, nullptr);
+    AddBindings<eUniformTexelBuffer>(bindings, binding, flags, entries.texel_buffers);
-        }
+    AddBindings<eCombinedImageSampler>(bindings, binding, flags, entries.samplers);
-    };
+    AddBindings<eStorageImage>(bindings, binding, flags, entries.images);
-    AddBindings(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
-    AddBindings(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
-    AddBindings(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
-    AddBindings(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
-    AddBindings(vk::DescriptorType::eStorageImage, entries.images.size());
    return binding;
 }
@@ -361,32 +377,45 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
    return {std::move(program), std::move(bindings)};
 }
-void FillDescriptorUpdateTemplateEntries(
+template <vk::DescriptorType descriptor_type, class Container>
-    const VKDevice& device, const ShaderEntries& entries, u32& binding, u32& offset,
+void AddEntry(std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries, u32& binding,
-    std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) {
+              u32& offset, const Container& container) {
-    static constexpr auto entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry));
+    static constexpr u32 entry_size = static_cast<u32>(sizeof(DescriptorUpdateEntry));
-    const auto AddEntry = [&](vk::DescriptorType descriptor_type, std::size_t count_) {
+    const u32 count = static_cast<u32>(std::size(container));
-        const u32 count = static_cast<u32>(count_);
-        if (descriptor_type == vk::DescriptorType::eUniformTexelBuffer &&
+    if constexpr (descriptor_type == eCombinedImageSampler) {
-            device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) {
+        for (u32 i = 0; i < count; ++i) {
-            // Nvidia has a bug where updating multiple uniform texels at once causes the driver to
+            const u32 num_samplers = container[i].Size();
-            // crash.
+            template_entries.emplace_back(binding, 0, num_samplers, descriptor_type, offset,
-            for (u32 i = 0; i < count; ++i) {
+                                          entry_size);
-                template_entries.emplace_back(binding + i, 0, 1, descriptor_type,
+            ++binding;
-                                              offset + i * entry_size, entry_size);
+            offset += num_samplers * entry_size;
-            }
-        } else if (count != 0) {
-            template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size);
        }
-        offset += count * entry_size;
+        return;
-        binding += count;
+    }
-    };
-    AddEntry(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
+    if constexpr (descriptor_type == eUniformTexelBuffer) {
-    AddEntry(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
+        // Nvidia has a bug where updating multiple uniform texels at once causes the driver to
-    AddEntry(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
+        // crash.
-    AddEntry(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
+        for (u32 i = 0; i < count; ++i) {
-    AddEntry(vk::DescriptorType::eStorageImage, entries.images.size());
+            template_entries.emplace_back(binding + i, 0, 1, descriptor_type,
+                                          offset + i * entry_size, entry_size);
+        }
+    } else if (count > 0) {
+        template_entries.emplace_back(binding, 0, count, descriptor_type, offset, entry_size);
+    }
+    offset += count * entry_size;
+    binding += count;
+}
+void FillDescriptorUpdateTemplateEntries(
+    const ShaderEntries& entries, u32& binding, u32& offset,
+    std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries) {
+    AddEntry<eUniformBuffer>(template_entries, offset, binding, entries.const_buffers);
+    AddEntry<eStorageBuffer>(template_entries, offset, binding, entries.global_buffers);
+    AddEntry<eUniformTexelBuffer>(template_entries, offset, binding, entries.texel_buffers);
+    AddEntry<eCombinedImageSampler>(template_entries, offset, binding, entries.samplers);
+    AddEntry<eStorageImage>(template_entries, offset, binding, entries.images);
 }
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
index 8678fc9c3..92a670cc7 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
@@ -194,7 +194,7 @@ private:
 };
 void FillDescriptorUpdateTemplateEntries(
-    const VKDevice& device, const ShaderEntries& entries, u32& binding, u32& offset,
+    const ShaderEntries& entries, u32& binding, u32& offset,
    std::vector<vk::DescriptorUpdateTemplateEntry>& template_entries);
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index 3bf86da87..3fe28c204 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -105,17 +105,20 @@ void TransitionImages(const std::vector<ImageView>& views, vk::PipelineStageFlag
 template <typename Engine, typename Entry>
 Tegra::Texture::FullTextureInfo GetTextureInfo(const Engine& engine, const Entry& entry,
-                                               std::size_t stage) {
+                                               std::size_t stage, std::size_t index = 0) {
    const auto stage_type = static_cast<Tegra::Engines::ShaderType>(stage);
    if (entry.IsBindless()) {
        const Tegra::Texture::TextureHandle tex_handle =
            engine.AccessConstBuffer32(stage_type, entry.GetBuffer(), entry.GetOffset());
        return engine.GetTextureInfo(tex_handle);
    }
+    const auto& gpu_profile = engine.AccessGuestDriverProfile();
+    const u32 entry_offset = static_cast<u32>(index * gpu_profile.GetTextureHandlerSize());
+    const u32 offset = entry.GetOffset() + entry_offset;
    if constexpr (std::is_same_v<Engine, Tegra::Engines::Maxwell3D>) {
-        return engine.GetStageTexture(stage_type, entry.GetOffset());
+        return engine.GetStageTexture(stage_type, offset);
    } else {
-        return engine.GetTexture(entry.GetOffset());
+        return engine.GetTexture(offset);
    }
 }
@@ -836,8 +839,10 @@ void RasterizerVulkan::SetupGraphicsTextures(const ShaderEntries& entries, std::
    MICROPROFILE_SCOPE(Vulkan_Textures);
    const auto& gpu = system.GPU().Maxwell3D();
    for (const auto& entry : entries.samplers) {
-        const auto texture = GetTextureInfo(gpu, entry, stage);
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
-        SetupTexture(texture, entry);
+            const auto texture = GetTextureInfo(gpu, entry, stage, i);
+            SetupTexture(texture, entry);
+        }
    }
 }
@@ -886,8 +891,10 @@ void RasterizerVulkan::SetupComputeTextures(const ShaderEntries& entries) {
    MICROPROFILE_SCOPE(Vulkan_Textures);
    const auto& gpu = system.GPU().KeplerCompute();
    for (const auto& entry : entries.samplers) {
-        const auto texture = GetTextureInfo(gpu, entry, ComputeShaderIndex);
+        for (std::size_t i = 0; i < entry.Size(); ++i) {
-        SetupTexture(texture, entry);
+            const auto texture = GetTextureInfo(gpu, entry, ComputeShaderIndex, i);
+            SetupTexture(texture, entry);
+        }
    }
 }
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 2da622d15..cfcca5af0 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -69,8 +69,9 @@ struct TexelBuffer {
 struct SampledImage {
    Id image_type{};
-    Id sampled_image_type{};
+    Id sampler_type{};
-    Id sampler{};
+    Id sampler_pointer_type{};
+    Id variable{};
 };
 struct StorageImage {
@@ -833,16 +834,20 @@ private:
            constexpr int sampled = 1;
            constexpr auto format = spv::ImageFormat::Unknown;
            const Id image_type = TypeImage(t_float, dim, depth, arrayed, ms, sampled, format);
-            const Id sampled_image_type = TypeSampledImage(image_type);
+            const Id sampler_type = TypeSampledImage(image_type);
-            const Id pointer_type =
+            const Id sampler_pointer_type =
-                TypePointer(spv::StorageClass::UniformConstant, sampled_image_type);
+                TypePointer(spv::StorageClass::UniformConstant, sampler_type);
+            const Id type = sampler.IsIndexed()
+                                ? TypeArray(sampler_type, Constant(t_uint, sampler.Size()))
+                                : sampler_type;
+            const Id pointer_type = TypePointer(spv::StorageClass::UniformConstant, type);
            const Id id = OpVariable(pointer_type, spv::StorageClass::UniformConstant);
            AddGlobalVariable(Name(id, fmt::format("sampler_{}", sampler.GetIndex())));
            Decorate(id, spv::Decoration::Binding, binding++);
            Decorate(id, spv::Decoration::DescriptorSet, DESCRIPTOR_SET);
-            sampled_images.emplace(sampler.GetIndex(),
+            sampled_images.emplace(sampler.GetIndex(), SampledImage{image_type, sampler_type,
-                                   SampledImage{image_type, sampled_image_type, id});
+                                                                    sampler_pointer_type, id});
        }
        return binding;
    }
@@ -1525,7 +1530,12 @@ private:
        ASSERT(!meta.sampler.IsBuffer());
        const auto& entry = sampled_images.at(meta.sampler.GetIndex());
-        return OpLoad(entry.sampled_image_type, entry.sampler);
+        Id sampler = entry.variable;
+        if (meta.sampler.IsIndexed()) {
+            const Id index = AsInt(Visit(meta.index));
+            sampler = OpAccessChain(entry.sampler_pointer_type, sampler, index);
+        }
+        return OpLoad(entry.sampler_type, sampler);
    }
    Id GetTextureImage(Operation operation) {
@@ -2211,16 +2221,14 @@ private:
        switch (specialization.attribute_types.at(location)) {
        case Maxwell::VertexAttribute::Type::SignedNorm:
        case Maxwell::VertexAttribute::Type::UnsignedNorm:
+        case Maxwell::VertexAttribute::Type::UnsignedScaled:
+        case Maxwell::VertexAttribute::Type::SignedScaled:
        case Maxwell::VertexAttribute::Type::Float:
            return {Type::Float, t_in_float, t_in_float4};
        case Maxwell::VertexAttribute::Type::SignedInt:
            return {Type::Int, t_in_int, t_in_int4};
        case Maxwell::VertexAttribute::Type::UnsignedInt:
            return {Type::Uint, t_in_uint, t_in_uint4};
-        case Maxwell::VertexAttribute::Type::UnsignedScaled:
-        case Maxwell::VertexAttribute::Type::SignedScaled:
-            UNIMPLEMENTED();
-            return {Type::Float, t_in_float, t_in_float4};
        default:
            UNREACHABLE();
            return {Type::Float, t_in_float, t_in_float4};
diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp
index 90240c765..478394682 100644
--- a/src/video_core/shader/decode/arithmetic.cpp
+++ b/src/video_core/shader/decode/arithmetic.cpp
@@ -53,29 +53,24 @@ u32 ShaderIR::DecodeArithmetic(NodeBlock& bb, u32 pc) {
        op_b = GetOperandAbsNegFloat(op_b, false, instr.fmul.negate_b);
-        // TODO(Rodrigo): Should precise be used when there's a postfactor?
+        static constexpr std::array FmulPostFactor = {
-        Node value = Operation(OperationCode::FMul, PRECISE, op_a, op_b);
+            1.000f, // None
+            0.500f, // Divide 2
+            0.250f, // Divide 4
+            0.125f, // Divide 8
+            8.000f, // Mul 8
+            4.000f, // Mul 4
+            2.000f, // Mul 2
+        };
        if (instr.fmul.postfactor != 0) {
-            auto postfactor = static_cast<s32>(instr.fmul.postfactor);
+            op_a = Operation(OperationCode::FMul, NO_PRECISE, op_a,
+                             Immediate(FmulPostFactor[instr.fmul.postfactor]));
-            // Postfactor encoded as 3-bit 1's complement in instruction, interpreted with below
-            // logic.
-            if (postfactor >= 4) {
-                postfactor = 7 - postfactor;
-            } else {
-                postfactor = 0 - postfactor;
-            }
-            if (postfactor > 0) {
-                value = Operation(OperationCode::FMul, NO_PRECISE, value,
-                                  Immediate(static_cast<f32>(1 << postfactor)));
-            } else {
-                value = Operation(OperationCode::FDiv, NO_PRECISE, value,
-                                  Immediate(static_cast<f32>(1 << -postfactor)));
-            }
        }
+        // TODO(Rodrigo): Should precise be used when there's a postfactor?
+        Node value = Operation(OperationCode::FMul, PRECISE, op_a, op_b);
        value = GetSaturatedFloat(value, instr.alu.saturate_d);
        SetInternalFlagsFromFloat(bb, value, instr.generates_cc);
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index 21366869d..2fe787d6f 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -293,44 +293,66 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
 void ShaderIR::WriteLop3Instruction(NodeBlock& bb, Register dest, Node op_a, Node op_b, Node op_c,
                                    Node imm_lut, bool sets_cc) {
-    constexpr u32 lop_iterations = 32;
+    const Node lop3_fast = [&](const Node na, const Node nb, const Node nc, const Node ttbl) {
-    const Node one = Immediate(1);
+        Node value = Immediate(0);
-    const Node two = Immediate(2);
+        const ImmediateNode imm = std::get<ImmediateNode>(*ttbl);
+        if (imm.GetValue() & 0x01) {
-    Node value;
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
-    for (u32 i = 0; i < lop_iterations; ++i) {
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
-        const Node shift_amount = Immediate(i);
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, b);
-        const Node a = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_c, shift_amount);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
-        const Node pack_0 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, one);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
-        const Node b = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_b, shift_amount);
-        const Node c = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, b, one);
-        const Node pack_1 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, c, one);
-        const Node d = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_a, shift_amount);
-        const Node e = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, d, one);
-        const Node pack_2 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, e, two);
-        const Node pack_01 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_0, pack_1);
-        const Node pack_012 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_01, pack_2);
-        const Node shifted_bit =
-            Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, imm_lut, pack_012);
-        const Node bit = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, shifted_bit, one);
-        const Node right =
-            Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, bit, shift_amount);
-        if (i > 0) {
-            value = Operation(OperationCode::IBitwiseOr, NO_PRECISE, value, right);
-        } else {
-            value = right;
        }
-    }
+        if (imm.GetValue() & 0x02) {
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, b);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x04) {
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x08) {
+            const Node a = Operation(OperationCode::IBitwiseNot, na);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x10) {
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, b);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x20) {
+            const Node b = Operation(OperationCode::IBitwiseNot, nb);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, b);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x40) {
+            const Node c = Operation(OperationCode::IBitwiseNot, nc);
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, c);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        if (imm.GetValue() & 0x80) {
+            Node r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, na, nb);
+            r = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, r, nc);
+            value = Operation(OperationCode::IBitwiseOr, value, r);
+        }
+        return value;
+    }(op_a, op_b, op_c, imm_lut);
-    SetInternalFlagsFromInteger(bb, value, sets_cc);
+    SetInternalFlagsFromInteger(bb, lop3_fast, sets_cc);
-    SetRegister(bb, dest, value);
+    SetRegister(bb, dest, lop3_fast);
 }
 } // namespace VideoCommon::Shader
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index a0a7b9111..a1828546e 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -299,7 +299,7 @@ private:
    u32 index{};  ///< Emulated index given for the this sampler.
    u32 offset{}; ///< Offset in the const buffer from where the sampler is being read.
    u32 buffer{}; ///< Buffer where the bindless sampler is being read (unused on bound samplers).
-    u32 size{};   ///< Size of the sampler if indexed.
+    u32 size{1};  ///< Size of the sampler.
    Tegra::Shader::TextureType type{}; ///< The type used to sample this texture (Texture2D, etc)
    bool is_array{};    ///< Whether the texture is being sampled as an array texture or not.
diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp
index face8c943..15e22b9fa 100644
--- a/src/video_core/shader/track.cpp
+++ b/src/video_core/shader/track.cpp
@@ -157,13 +157,21 @@ std::tuple<Node, u32, u32> ShaderIR::TrackCbuf(Node tracked, const NodeBlock& co
        if (gpr->GetIndex() == Tegra::Shader::Register::ZeroIndex) {
            return {};
        }
-        // Reduce the cursor in one to avoid infinite loops when the instruction sets the same
+        s64 current_cursor = cursor;
-        // register that it uses as operand
+        while (current_cursor > 0) {
-        const auto [source, new_cursor] = TrackRegister(gpr, code, cursor - 1);
+            // Reduce the cursor in one to avoid infinite loops when the instruction sets the same
-        if (!source) {
+            // register that it uses as operand
-            return {};
+            const auto [source, new_cursor] = TrackRegister(gpr, code, current_cursor - 1);
+            current_cursor = new_cursor;
+            if (!source) {
+                continue;
+            }
+            const auto [base_address, index, offset] = TrackCbuf(source, code, current_cursor);
+            if (base_address != nullptr) {
+                return {base_address, index, offset};
+            }
        }
-        return TrackCbuf(source, code, new_cursor);
+        return {};
    }
    if (const auto operation = std::get_if<OperationNode>(&*tracked)) {
        for (std::size_t i = operation->GetOperandsCount(); i > 0; --i) {
diff --git a/src/yuzu/bootmanager.cpp b/src/yuzu/bootmanager.cpp
index 55a37fffa..c3dbb1a88 100644
--- a/src/yuzu/bootmanager.cpp
+++ b/src/yuzu/bootmanager.cpp
@@ -9,6 +9,9 @@
 #include <QKeyEvent>
 #include <QMessageBox>
 #include <QOffscreenSurface>
+#include <QOpenGLContext>
+#include <QOpenGLFunctions>
+#include <QOpenGLFunctions_4_3_Core>
 #include <QOpenGLWindow>
 #include <QPainter>
 #include <QScreen>
@@ -23,9 +26,10 @@
 #include "common/assert.h"
 #include "common/microprofile.h"
 #include "common/scm_rev.h"
+#include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/frontend/framebuffer_layout.h"
-#include "core/frontend/scope_acquire_window_context.h"
+#include "core/frontend/scope_acquire_context.h"
 #include "core/settings.h"
 #include "input_common/keyboard.h"
 #include "input_common/main.h"
@@ -35,15 +39,27 @@
 #include "yuzu/bootmanager.h"
 #include "yuzu/main.h"
-EmuThread::EmuThread(GRenderWindow* render_window) : render_window(render_window) {}
+EmuThread::EmuThread(GRenderWindow& window)
+    : shared_context{window.CreateSharedContext()},
+      context{(Settings::values.use_asynchronous_gpu_emulation && shared_context) ? *shared_context
+                                                                                  : window} {}
 EmuThread::~EmuThread() = default;
-void EmuThread::run() {
+static GMainWindow* GetMainWindow() {
-    render_window->MakeCurrent();
+    for (QWidget* w : qApp->topLevelWidgets()) {
+        if (GMainWindow* main = qobject_cast<GMainWindow*>(w)) {
+            return main;
+        }
+    }
+    return nullptr;
+}
+void EmuThread::run() {
    MicroProfileOnThreadCreate("EmuThread");
+    Core::Frontend::ScopeAcquireContext acquire_context{context};
    emit LoadProgress(VideoCore::LoadCallbackStage::Prepare, 0, 0);
    Core::System::GetInstance().Renderer().Rasterizer().LoadDiskResources(
@@ -53,11 +69,6 @@ void EmuThread::run() {
    emit LoadProgress(VideoCore::LoadCallbackStage::Complete, 0, 0);
-    if (Settings::values.use_asynchronous_gpu_emulation) {
-        // Release OpenGL context for the GPU thread
-        render_window->DoneCurrent();
-    }
    // Holds whether the cpu was running during the last iteration,
    // so that the DebugModeLeft signal can be emitted before the
    // next execution step
@@ -98,190 +109,202 @@ void EmuThread::run() {
 #if MICROPROFILE_ENABLED
    MicroProfileOnThreadExit();
 #endif
-    render_window->moveContext();
 }
 class GGLContext : public Core::Frontend::GraphicsContext {
 public:
-    explicit GGLContext(QOpenGLContext* shared_context) : shared_context{shared_context} {
+    explicit GGLContext(QOpenGLContext* shared_context)
-        context.setFormat(shared_context->format());
+        : context(new QOpenGLContext(shared_context->parent())),
-        context.setShareContext(shared_context);
+          surface(new QOffscreenSurface(nullptr)) {
-        context.create();
+        // disable vsync for any shared contexts
+        auto format = shared_context->format();
+        format.setSwapInterval(0);
+        context->setShareContext(shared_context);
+        context->setFormat(format);
+        context->create();
+        surface->setParent(shared_context->parent());
+        surface->setFormat(format);
+        surface->create();
    }
    void MakeCurrent() override {
-        context.makeCurrent(shared_context->surface());
+        context->makeCurrent(surface);
    }
    void DoneCurrent() override {
-        context.doneCurrent();
+        context->doneCurrent();
    }
-    void SwapBuffers() override {}
 private:
-    QOpenGLContext* shared_context;
+    QOpenGLContext* context;
-    QOpenGLContext context;
+    QOffscreenSurface* surface;
 };
-class GWidgetInternal : public QWindow {
+class ChildRenderWindow : public QWindow {
 public:
-    GWidgetInternal(GRenderWindow* parent) : parent(parent) {}
+    ChildRenderWindow(QWindow* parent, QWidget* event_handler)
-    virtual ~GWidgetInternal() = default;
+        : QWindow{parent}, event_handler{event_handler} {}
-    void resizeEvent(QResizeEvent* ev) override {
+    virtual ~ChildRenderWindow() = default;
-        parent->OnClientAreaResized(ev->size().width(), ev->size().height());
-        parent->OnFramebufferSizeChanged();
-    }
-    void keyPressEvent(QKeyEvent* event) override {
+    virtual void Present() = 0;
-        InputCommon::GetKeyboard()->PressKey(event->key());
-    }
-    void keyReleaseEvent(QKeyEvent* event) override {
+protected:
-        InputCommon::GetKeyboard()->ReleaseKey(event->key());
+    bool event(QEvent* event) override {
+        switch (event->type()) {
+        case QEvent::UpdateRequest:
+            Present();
+            return true;
+        case QEvent::MouseButtonPress:
+        case QEvent::MouseButtonRelease:
+        case QEvent::MouseButtonDblClick:
+        case QEvent::MouseMove:
+        case QEvent::KeyPress:
+        case QEvent::KeyRelease:
+        case QEvent::FocusIn:
+        case QEvent::FocusOut:
+        case QEvent::FocusAboutToChange:
+        case QEvent::Enter:
+        case QEvent::Leave:
+        case QEvent::Wheel:
+        case QEvent::TabletMove:
+        case QEvent::TabletPress:
+        case QEvent::TabletRelease:
+        case QEvent::TabletEnterProximity:
+        case QEvent::TabletLeaveProximity:
+        case QEvent::TouchBegin:
+        case QEvent::TouchUpdate:
+        case QEvent::TouchEnd:
+        case QEvent::InputMethodQuery:
+        case QEvent::TouchCancel:
+            return QCoreApplication::sendEvent(event_handler, event);
+        case QEvent::Drop:
+            GetMainWindow()->DropAction(static_cast<QDropEvent*>(event));
+            return true;
+        case QEvent::DragResponse:
+        case QEvent::DragEnter:
+        case QEvent::DragLeave:
+        case QEvent::DragMove:
+            GetMainWindow()->AcceptDropEvent(static_cast<QDropEvent*>(event));
+            return true;
+        default:
+            return QWindow::event(event);
+        }
    }
-    void mousePressEvent(QMouseEvent* event) override {
+    void exposeEvent(QExposeEvent* event) override {
-        if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        QWindow::requestUpdate();
-            return; // touch input is handled in TouchBeginEvent
+        QWindow::exposeEvent(event);
-        const auto pos{event->pos()};
-        if (event->button() == Qt::LeftButton) {
-            const auto [x, y] = parent->ScaleTouch(pos);
-            parent->TouchPressed(x, y);
-        } else if (event->button() == Qt::RightButton) {
-            InputCommon::GetMotionEmu()->BeginTilt(pos.x(), pos.y());
-        }
    }
-    void mouseMoveEvent(QMouseEvent* event) override {
+private:
-        if (event->source() == Qt::MouseEventSynthesizedBySystem)
+    QWidget* event_handler{};
-            return; // touch input is handled in TouchUpdateEvent
+};
-        const auto pos{event->pos()};
+class OpenGLWindow final : public ChildRenderWindow {
-        const auto [x, y] = parent->ScaleTouch(pos);
+public:
-        parent->TouchMoved(x, y);
+    OpenGLWindow(QWindow* parent, QWidget* event_handler, QOpenGLContext* shared_context)
-        InputCommon::GetMotionEmu()->Tilt(pos.x(), pos.y());
+        : ChildRenderWindow{parent, event_handler},
-    }
+          context(new QOpenGLContext(shared_context->parent())) {
-    void mouseReleaseEvent(QMouseEvent* event) override {
+        // disable vsync for any shared contexts
-        if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        auto format = shared_context->format();
-            return; // touch input is handled in TouchEndEvent
+        format.setSwapInterval(Settings::values.use_vsync ? 1 : 0);
+        this->setFormat(format);
-        if (event->button() == Qt::LeftButton)
+        context->setShareContext(shared_context);
-            parent->TouchReleased();
+        context->setScreen(this->screen());
-        else if (event->button() == Qt::RightButton)
+        context->setFormat(format);
-            InputCommon::GetMotionEmu()->EndTilt();
+        context->create();
-    }
-    void DisablePainting() {
+        setSurfaceType(QWindow::OpenGLSurface);
-        do_painting = false;
-    }
-    void EnablePainting() {
+        // TODO: One of these flags might be interesting: WA_OpaquePaintEvent, WA_NoBackground,
-        do_painting = true;
+        // WA_DontShowOnScreen, WA_DeleteOnClose
    }
-    std::pair<unsigned, unsigned> GetSize() const {
+    ~OpenGLWindow() override {
-        return std::make_pair(width(), height());
+        context->doneCurrent();
    }
-protected:
+    void Present() override {
-    bool IsPaintingEnabled() const {
+        if (!isExposed()) {
-        return do_painting;
+            return;
+        }
+        context->makeCurrent(this);
+        Core::System::GetInstance().Renderer().TryPresent(100);
+        context->swapBuffers(this);
+        auto f = context->versionFunctions<QOpenGLFunctions_4_3_Core>();
+        f->glFinish();
+        QWindow::requestUpdate();
    }
 private:
-    GRenderWindow* parent;
+    QOpenGLContext* context{};
-    bool do_painting = false;
-};
-// This class overrides paintEvent and resizeEvent to prevent the GUI thread from stealing GL
-// context.
-// The corresponding functionality is handled in EmuThread instead
-class GGLWidgetInternal final : public GWidgetInternal, public QOpenGLWindow {
-public:
-    GGLWidgetInternal(GRenderWindow* parent, QOpenGLContext* shared_context)
-        : GWidgetInternal(parent), QOpenGLWindow(shared_context) {}
-    ~GGLWidgetInternal() override = default;
-    void paintEvent(QPaintEvent* ev) override {
-        if (IsPaintingEnabled()) {
-            QPainter painter(this);
-        }
-    }
 };
 #ifdef HAS_VULKAN
-class GVKWidgetInternal final : public GWidgetInternal {
+class VulkanWindow final : public ChildRenderWindow {
 public:
-    GVKWidgetInternal(GRenderWindow* parent, QVulkanInstance* instance) : GWidgetInternal(parent) {
+    VulkanWindow(QWindow* parent, QWidget* event_handler, QVulkanInstance* instance)
+        : ChildRenderWindow{parent, event_handler} {
        setSurfaceType(QSurface::SurfaceType::VulkanSurface);
        setVulkanInstance(instance);
    }
-    ~GVKWidgetInternal() override = default;
+    ~VulkanWindow() override = default;
+    void Present() override {
+        // TODO(bunnei): ImplementMe
+    }
+private:
+    QWidget* event_handler{};
 };
 #endif
-GRenderWindow::GRenderWindow(GMainWindow* parent, EmuThread* emu_thread)
+GRenderWindow::GRenderWindow(QWidget* parent_, EmuThread* emu_thread)
-    : QWidget(parent), emu_thread(emu_thread) {
+    : QWidget(parent_), emu_thread(emu_thread) {
    setWindowTitle(QStringLiteral("yuzu %1 | %2-%3")
                       .arg(QString::fromUtf8(Common::g_build_name),
                            QString::fromUtf8(Common::g_scm_branch),
                            QString::fromUtf8(Common::g_scm_desc)));
    setAttribute(Qt::WA_AcceptTouchEvents);
+    auto layout = new QHBoxLayout(this);
+    layout->setMargin(0);
+    setLayout(layout);
    InputCommon::Init();
+    GMainWindow* parent = GetMainWindow();
    connect(this, &GRenderWindow::FirstFrameDisplayed, parent, &GMainWindow::OnLoadComplete);
 }
 GRenderWindow::~GRenderWindow() {
    InputCommon::Shutdown();
-    // Avoid an unordered destruction that generates a segfault
-    delete child;
 }
-void GRenderWindow::moveContext() {
+void GRenderWindow::MakeCurrent() {
-    if (!context) {
+    if (core_context) {
-        return;
+        core_context->MakeCurrent();
    }
-    DoneCurrent();
-    // If the thread started running, move the GL Context to the new thread. Otherwise, move it
-    // back.
-    auto thread = (QThread::currentThread() == qApp->thread() && emu_thread != nullptr)
-                      ? emu_thread
-                      : qApp->thread();
-    context->moveToThread(thread);
 }
-void GRenderWindow::SwapBuffers() {
+void GRenderWindow::DoneCurrent() {
-    if (context) {
+    if (core_context) {
-        context->swapBuffers(child);
+        core_context->DoneCurrent();
    }
+}
+void GRenderWindow::PollEvents() {
    if (!first_frame) {
        first_frame = true;
        emit FirstFrameDisplayed();
    }
 }
-void GRenderWindow::MakeCurrent() {
-    if (context) {
-        context->makeCurrent(child);
-    }
-}
-void GRenderWindow::DoneCurrent() {
-    if (context) {
-        context->doneCurrent();
-    }
-}
-void GRenderWindow::PollEvents() {}
 bool GRenderWindow::IsShown() const {
    return !isMinimized();
 }
@@ -291,7 +314,7 @@ void GRenderWindow::RetrieveVulkanHandlers(void* get_instance_proc_addr, void* i
 #ifdef HAS_VULKAN
    const auto instance_proc_addr = vk_instance->getInstanceProcAddr("vkGetInstanceProcAddr");
    const VkInstance instance_copy = vk_instance->vkInstance();
-    const VkSurfaceKHR surface_copy = vk_instance->surfaceForWindow(child);
+    const VkSurfaceKHR surface_copy = vk_instance->surfaceForWindow(child_window);
    std::memcpy(get_instance_proc_addr, &instance_proc_addr, sizeof(instance_proc_addr));
    std::memcpy(instance, &instance_copy, sizeof(instance_copy));
@@ -309,21 +332,10 @@ void GRenderWindow::RetrieveVulkanHandlers(void* get_instance_proc_addr, void* i
 void GRenderWindow::OnFramebufferSizeChanged() {
    // Screen changes potentially incur a change in screen DPI, hence we should update the
    // framebuffer size
-    const qreal pixelRatio{GetWindowPixelRatio()};
+    const qreal pixel_ratio = windowPixelRatio();
-    const auto size{child->GetSize()};
+    const u32 width = this->width() * pixel_ratio;
-    UpdateCurrentFramebufferLayout(size.first * pixelRatio, size.second * pixelRatio);
+    const u32 height = this->height() * pixel_ratio;
-}
+    UpdateCurrentFramebufferLayout(width, height);
-void GRenderWindow::ForwardKeyPressEvent(QKeyEvent* event) {
-    if (child) {
-        child->keyPressEvent(event);
-    }
-}
-void GRenderWindow::ForwardKeyReleaseEvent(QKeyEvent* event) {
-    if (child) {
-        child->keyReleaseEvent(event);
-    }
 }
 void GRenderWindow::BackupGeometry() {
@@ -351,13 +363,12 @@ QByteArray GRenderWindow::saveGeometry() {
    return geometry;
 }
-qreal GRenderWindow::GetWindowPixelRatio() const {
+qreal GRenderWindow::windowPixelRatio() const {
-    // windowHandle() might not be accessible until the window is displayed to screen.
+    return devicePixelRatio();
-    return windowHandle() ? windowHandle()->screen()->devicePixelRatio() : 1.0f;
 }
 std::pair<u32, u32> GRenderWindow::ScaleTouch(const QPointF pos) const {
-    const qreal pixel_ratio{GetWindowPixelRatio()};
+    const qreal pixel_ratio = windowPixelRatio();
    return {static_cast<u32>(std::max(std::round(pos.x() * pixel_ratio), qreal{0.0})),
            static_cast<u32>(std::max(std::round(pos.y() * pixel_ratio), qreal{0.0}))};
 }
@@ -367,6 +378,47 @@ void GRenderWindow::closeEvent(QCloseEvent* event) {
    QWidget::closeEvent(event);
 }
+void GRenderWindow::keyPressEvent(QKeyEvent* event) {
+    InputCommon::GetKeyboard()->PressKey(event->key());
+}
+void GRenderWindow::keyReleaseEvent(QKeyEvent* event) {
+    InputCommon::GetKeyboard()->ReleaseKey(event->key());
+}
+void GRenderWindow::mousePressEvent(QMouseEvent* event) {
+    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        return; // touch input is handled in TouchBeginEvent
+    auto pos = event->pos();
+    if (event->button() == Qt::LeftButton) {
+        const auto [x, y] = ScaleTouch(pos);
+        this->TouchPressed(x, y);
+    } else if (event->button() == Qt::RightButton) {
+        InputCommon::GetMotionEmu()->BeginTilt(pos.x(), pos.y());
+    }
+}
+void GRenderWindow::mouseMoveEvent(QMouseEvent* event) {
+    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        return; // touch input is handled in TouchUpdateEvent
+    auto pos = event->pos();
+    const auto [x, y] = ScaleTouch(pos);
+    this->TouchMoved(x, y);
+    InputCommon::GetMotionEmu()->Tilt(pos.x(), pos.y());
+}
+void GRenderWindow::mouseReleaseEvent(QMouseEvent* event) {
+    if (event->source() == Qt::MouseEventSynthesizedBySystem)
+        return; // touch input is handled in TouchEndEvent
+    if (event->button() == Qt::LeftButton)
+        this->TouchReleased();
+    else if (event->button() == Qt::RightButton)
+        InputCommon::GetMotionEmu()->EndTilt();
+}
 void GRenderWindow::TouchBeginEvent(const QTouchEvent* event) {
    // TouchBegin always has exactly one touch point, so take the .first()
    const auto [x, y] = ScaleTouch(event->touchPoints().first().pos());
@@ -415,26 +467,20 @@ void GRenderWindow::focusOutEvent(QFocusEvent* event) {
    InputCommon::GetKeyboard()->ReleaseAllKeys();
 }
-void GRenderWindow::OnClientAreaResized(u32 width, u32 height) {
+void GRenderWindow::resizeEvent(QResizeEvent* event) {
-    NotifyClientAreaSizeChanged(std::make_pair(width, height));
+    QWidget::resizeEvent(event);
+    OnFramebufferSizeChanged();
 }
 std::unique_ptr<Core::Frontend::GraphicsContext> GRenderWindow::CreateSharedContext() const {
-    return std::make_unique<GGLContext>(context.get());
+    if (Settings::values.renderer_backend == Settings::RendererBackend::OpenGL) {
+        return std::make_unique<GGLContext>(QOpenGLContext::globalShareContext());
+    }
+    return {};
 }
 bool GRenderWindow::InitRenderTarget() {
-    shared_context.reset();
+    ReleaseRenderTarget();
-    context.reset();
-    if (child) {
-        delete child;
-    }
-    if (container) {
-        delete container;
-    }
-    if (layout()) {
-        delete layout();
-    }
    first_frame = false;
@@ -451,13 +497,6 @@ bool GRenderWindow::InitRenderTarget() {
        break;
    }
-    container = QWidget::createWindowContainer(child, this);
-    QBoxLayout* layout = new QHBoxLayout(this);
-    layout->addWidget(container);
-    layout->setMargin(0);
-    setLayout(layout);
    // Reset minimum required size to avoid resizing issues on the main window after restarting.
    setMinimumSize(1, 1);
@@ -467,14 +506,9 @@ bool GRenderWindow::InitRenderTarget() {
    hide();
    resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
-    child->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
-    container->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
    OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
    OnFramebufferSizeChanged();
-    NotifyClientAreaSizeChanged(child->GetSize());
    BackupGeometry();
    if (Settings::values.renderer_backend == Settings::RendererBackend::OpenGL) {
@@ -486,6 +520,14 @@ bool GRenderWindow::InitRenderTarget() {
    return true;
 }
+void GRenderWindow::ReleaseRenderTarget() {
+    if (child_widget) {
+        layout()->removeWidget(child_widget);
+        delete child_widget;
+        child_widget = nullptr;
+    }
+}
 void GRenderWindow::CaptureScreenshot(u32 res_scale, const QString& screenshot_path) {
    auto& renderer = Core::System::GetInstance().Renderer();
@@ -521,16 +563,19 @@ bool GRenderWindow::InitializeOpenGL() {
    fmt.setOption(QSurfaceFormat::FormatOption::DeprecatedFunctions);
    // TODO: expose a setting for buffer value (ie default/single/double/triple)
    fmt.setSwapBehavior(QSurfaceFormat::DefaultSwapBehavior);
-    shared_context = std::make_unique<QOpenGLContext>();
+    fmt.setSwapInterval(0);
-    shared_context->setFormat(fmt);
+    QSurfaceFormat::setDefaultFormat(fmt);
-    shared_context->create();
-    context = std::make_unique<QOpenGLContext>();
+    GMainWindow* parent = GetMainWindow();
-    context->setShareContext(shared_context.get());
+    QWindow* parent_win_handle = parent ? parent->windowHandle() : nullptr;
-    context->setFormat(fmt);
+    child_window = new OpenGLWindow(parent_win_handle, this, QOpenGLContext::globalShareContext());
-    context->create();
+    child_window->create();
-    fmt.setSwapInterval(false);
+    child_widget = createWindowContainer(child_window, this);
+    child_widget->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
-    child = new GGLWidgetInternal(this, shared_context.get());
+    layout()->addWidget(child_widget);
+    core_context = CreateSharedContext();
    return true;
 }
@@ -559,7 +604,14 @@ bool GRenderWindow::InitializeVulkan() {
        return false;
    }
-    child = new GVKWidgetInternal(this, vk_instance.get());
+    GMainWindow* parent = GetMainWindow();
+    QWindow* parent_win_handle = parent ? parent->windowHandle() : nullptr;
+    child_window = new VulkanWindow(parent_win_handle, this, vk_instance.get());
+    child_window->create();
+    child_widget = createWindowContainer(child_window, this);
+    child_widget->resize(Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height);
+    layout()->addWidget(child_widget);
    return true;
 #else
    QMessageBox::critical(this, tr("Vulkan not available!"),
@@ -569,7 +621,7 @@ bool GRenderWindow::InitializeVulkan() {
 }
 bool GRenderWindow::LoadOpenGL() {
-    Core::Frontend::ScopeAcquireWindowContext acquire_context{*this};
+    Core::Frontend::ScopeAcquireContext acquire_context{*this};
    if (!gladLoadGL()) {
        QMessageBox::critical(this, tr("Error while initializing OpenGL 4.3!"),
                              tr("Your GPU may not support OpenGL 4.3, or you do not have the "
@@ -621,12 +673,10 @@ QStringList GRenderWindow::GetUnsupportedGLExtensions() const {
 void GRenderWindow::OnEmulationStarting(EmuThread* emu_thread) {
    this->emu_thread = emu_thread;
-    child->DisablePainting();
 }
 void GRenderWindow::OnEmulationStopping() {
    emu_thread = nullptr;
-    child->EnablePainting();
 }
 void GRenderWindow::showEvent(QShowEvent* event) {
diff --git a/src/yuzu/bootmanager.h b/src/yuzu/bootmanager.h
index 71a2fa321..79b030304 100644
--- a/src/yuzu/bootmanager.h
+++ b/src/yuzu/bootmanager.h
@@ -11,11 +11,13 @@
 #include <QImage>
 #include <QThread>
 #include <QWidget>
+#include <QWindow>
 #include "common/thread.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
+class GRenderWindow;
 class QKeyEvent;
 class QScreen;
 class QTouchEvent;
@@ -26,14 +28,6 @@ class QOpenGLContext;
 class QVulkanInstance;
 #endif
-class GWidgetInternal;
-class GGLWidgetInternal;
-class GVKWidgetInternal;
-class GMainWindow;
-class GRenderWindow;
-class QSurface;
-class QOpenGLContext;
 namespace VideoCore {
 enum class LoadCallbackStage;
 }
@@ -42,7 +36,7 @@ class EmuThread final : public QThread {
    Q_OBJECT
 public:
-    explicit EmuThread(GRenderWindow* render_window);
+    explicit EmuThread(GRenderWindow& window);
    ~EmuThread() override;
    /**
@@ -96,7 +90,11 @@ private:
    std::mutex running_mutex;
    std::condition_variable running_cv;
-    GRenderWindow* render_window;
+    /// Only used in asynchronous GPU mode
+    std::unique_ptr<Core::Frontend::GraphicsContext> shared_context;
+    /// This is shared_context in asynchronous GPU mode, core_context in synchronous GPU mode
+    Core::Frontend::GraphicsContext& context;
 signals:
    /**
@@ -126,11 +124,10 @@ class GRenderWindow : public QWidget, public Core::Frontend::EmuWindow {
    Q_OBJECT
 public:
-    GRenderWindow(GMainWindow* parent, EmuThread* emu_thread);
+    GRenderWindow(QWidget* parent, EmuThread* emu_thread);
    ~GRenderWindow() override;
-    // EmuWindow implementation
+    // EmuWindow implementation.
-    void SwapBuffers() override;
    void MakeCurrent() override;
    void DoneCurrent() override;
    void PollEvents() override;
@@ -139,30 +136,36 @@ public:
                                void* surface) const override;
    std::unique_ptr<Core::Frontend::GraphicsContext> CreateSharedContext() const override;
-    void ForwardKeyPressEvent(QKeyEvent* event);
-    void ForwardKeyReleaseEvent(QKeyEvent* event);
    void BackupGeometry();
    void RestoreGeometry();
    void restoreGeometry(const QByteArray& geometry); // overridden
    QByteArray saveGeometry();                        // overridden
-    qreal GetWindowPixelRatio() const;
+    qreal windowPixelRatio() const;
-    std::pair<u32, u32> ScaleTouch(QPointF pos) const;
    void closeEvent(QCloseEvent* event) override;
+    void resizeEvent(QResizeEvent* event) override;
+    void keyPressEvent(QKeyEvent* event) override;
+    void keyReleaseEvent(QKeyEvent* event) override;
+    void mousePressEvent(QMouseEvent* event) override;
+    void mouseMoveEvent(QMouseEvent* event) override;
+    void mouseReleaseEvent(QMouseEvent* event) override;
    bool event(QEvent* event) override;
-    void focusOutEvent(QFocusEvent* event) override;
-    void OnClientAreaResized(u32 width, u32 height);
+    void focusOutEvent(QFocusEvent* event) override;
    bool InitRenderTarget();
+    /// Destroy the previous run's child_widget which should also destroy the child_window
+    void ReleaseRenderTarget();
    void CaptureScreenshot(u32 res_scale, const QString& screenshot_path);
 public slots:
-    void moveContext(); // overridden
    void OnEmulationStarting(EmuThread* emu_thread);
    void OnEmulationStopping();
    void OnFramebufferSizeChanged();
@@ -173,6 +176,7 @@ signals:
    void FirstFrameDisplayed();
 private:
+    std::pair<u32, u32> ScaleTouch(QPointF pos) const;
    void TouchBeginEvent(const QTouchEvent* event);
    void TouchUpdateEvent(const QTouchEvent* event);
    void TouchEndEvent();
@@ -184,15 +188,9 @@ private:
    bool LoadOpenGL();
    QStringList GetUnsupportedGLExtensions() const;
-    QWidget* container = nullptr;
-    GWidgetInternal* child = nullptr;
    EmuThread* emu_thread;
-    // Context that backs the GGLWidgetInternal (and will be used by core to render)
-    std::unique_ptr<QOpenGLContext> context;
+    std::unique_ptr<GraphicsContext> core_context;
-    // Context that will be shared between all newly created contexts. This should never be made
-    // current
-    std::unique_ptr<QOpenGLContext> shared_context;
 #ifdef HAS_VULKAN
    std::unique_ptr<QVulkanInstance> vk_instance;
@@ -202,6 +200,15 @@ private:
    QImage screenshot_image;
    QByteArray geometry;
+    /// Native window handle that backs this presentation widget
+    QWindow* child_window = nullptr;
+    /// In order to embed the window into GRenderWindow, you need to use createWindowContainer to
+    /// put the child_window into a widget then add it to the layout. This child_widget can be
+    /// parented to GRenderWindow and use Qt's lifetime system
+    QWidget* child_widget = nullptr;
    bool first_frame = false;
 protected:
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index 6209fff75..d0f574147 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -640,6 +640,7 @@ void Config::ReadRendererValues() {
        ReadSetting(QStringLiteral("use_accurate_gpu_emulation"), false).toBool();
    Settings::values.use_asynchronous_gpu_emulation =
        ReadSetting(QStringLiteral("use_asynchronous_gpu_emulation"), false).toBool();
+    Settings::values.use_vsync = ReadSetting(QStringLiteral("use_vsync"), true).toBool();
    Settings::values.force_30fps_mode =
        ReadSetting(QStringLiteral("force_30fps_mode"), false).toBool();
@@ -1074,6 +1075,7 @@ void Config::SaveRendererValues() {
                 Settings::values.use_accurate_gpu_emulation, false);
    WriteSetting(QStringLiteral("use_asynchronous_gpu_emulation"),
                 Settings::values.use_asynchronous_gpu_emulation, false);
+    WriteSetting(QStringLiteral("use_vsync"), Settings::values.use_vsync, true);
    WriteSetting(QStringLiteral("force_30fps_mode"), Settings::values.force_30fps_mode, false);
    // Cast to double because Qt's written float values are not human-readable
diff --git a/src/yuzu/configuration/configure_graphics.cpp b/src/yuzu/configuration/configure_graphics.cpp
index ea899c080..fe64c7d81 100644
--- a/src/yuzu/configuration/configure_graphics.cpp
+++ b/src/yuzu/configuration/configure_graphics.cpp
@@ -103,6 +103,8 @@ void ConfigureGraphics::SetConfiguration() {
    ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
    ui->use_asynchronous_gpu_emulation->setEnabled(runtime_lock);
    ui->use_asynchronous_gpu_emulation->setChecked(Settings::values.use_asynchronous_gpu_emulation);
+    ui->use_vsync->setEnabled(runtime_lock);
+    ui->use_vsync->setChecked(Settings::values.use_vsync);
    ui->force_30fps_mode->setEnabled(runtime_lock);
    ui->force_30fps_mode->setChecked(Settings::values.force_30fps_mode);
    UpdateBackgroundColorButton(QColor::fromRgbF(Settings::values.bg_red, Settings::values.bg_green,
@@ -120,6 +122,7 @@ void ConfigureGraphics::ApplyConfiguration() {
    Settings::values.use_accurate_gpu_emulation = ui->use_accurate_gpu_emulation->isChecked();
    Settings::values.use_asynchronous_gpu_emulation =
        ui->use_asynchronous_gpu_emulation->isChecked();
+    Settings::values.use_vsync = ui->use_vsync->isChecked();
    Settings::values.force_30fps_mode = ui->force_30fps_mode->isChecked();
    Settings::values.bg_red = static_cast<float>(bg_color.redF());
    Settings::values.bg_green = static_cast<float>(bg_color.greenF());
diff --git a/src/yuzu/configuration/configure_graphics.ui b/src/yuzu/configuration/configure_graphics.ui
index db60426ab..9acc7dd93 100644
--- a/src/yuzu/configuration/configure_graphics.ui
+++ b/src/yuzu/configuration/configure_graphics.ui
@@ -85,6 +85,16 @@
         </widget>
        </item>
        <item>
+         <widget class="QCheckBox" name="use_vsync">
+          <property name="toolTip">
+           <string>VSync prevents the screen from tearing, but some graphics cards have lower performance with VSync enabled. Keep it enabled if you don't notice a performance difference.</string>
+          </property>
+          <property name="text">
+           <string>Use VSync (OpenGL only)</string>
+          </property>
+         </widget>
+        </item>
+        <item>
         <widget class="QCheckBox" name="use_accurate_gpu_emulation">
          <property name="text">
           <string>Use accurate GPU emulation (slow)</string>
diff --git a/src/yuzu/debugger/wait_tree.cpp b/src/yuzu/debugger/wait_tree.cpp
index 3f1a94627..c1ea25fb8 100644
--- a/src/yuzu/debugger/wait_tree.cpp
+++ b/src/yuzu/debugger/wait_tree.cpp
@@ -116,7 +116,7 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeCallstack::GetChildren() cons
    constexpr std::size_t BaseRegister = 29;
    auto& memory = Core::System::GetInstance().Memory();
-    u64 base_pointer = thread.GetContext().cpu_registers[BaseRegister];
+    u64 base_pointer = thread.GetContext64().cpu_registers[BaseRegister];
    while (base_pointer != 0) {
        const u64 lr = memory.Read64(base_pointer + sizeof(u64));
@@ -240,7 +240,7 @@ QString WaitTreeThread::GetText() const {
        break;
    }
-    const auto& context = thread.GetContext();
+    const auto& context = thread.GetContext64();
    const QString pc_info = tr(" PC = 0x%1 LR = 0x%2")
                                .arg(context.pc, 8, 16, QLatin1Char{'0'})
                                .arg(context.cpu_registers[30], 8, 16, QLatin1Char{'0'});
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index 54ca2dc1d..47615adfe 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -20,7 +20,6 @@
 #include "core/file_sys/vfs.h"
 #include "core/file_sys/vfs_real.h"
 #include "core/frontend/applets/general_frontend.h"
-#include "core/frontend/scope_acquire_window_context.h"
 #include "core/hle/service/acc/profile_manager.h"
 #include "core/hle/service/am/applet_ae.h"
 #include "core/hle/service/am/applet_oe.h"
@@ -985,11 +984,8 @@ void GMainWindow::BootGame(const QString& filename) {
        return;
    // Create and start the emulation thread
-    emu_thread = std::make_unique<EmuThread>(render_window);
+    emu_thread = std::make_unique<EmuThread>(*render_window);
    emit EmulationStarting(emu_thread.get());
-    if (Settings::values.renderer_backend == Settings::RendererBackend::OpenGL) {
-        render_window->moveContext();
-    }
    emu_thread->start();
    connect(render_window, &GRenderWindow::Closed, this, &GMainWindow::OnStopGame);
@@ -1087,6 +1083,9 @@ void GMainWindow::ShutdownGame() {
    emulation_running = false;
    game_path.clear();
+    // When closing the game, destroy the GLWindow to clear the context after the game is closed
+    render_window->ReleaseRenderTarget();
 }
 void GMainWindow::StoreRecentFile(const QString& filename) {
@@ -2215,48 +2214,47 @@ void GMainWindow::closeEvent(QCloseEvent* event) {
    QWidget::closeEvent(event);
 }
-void GMainWindow::keyPressEvent(QKeyEvent* event) {
+static bool IsSingleFileDropEvent(const QMimeData* mime) {
-    if (render_window) {
+    return mime->hasUrls() && mime->urls().length() == 1;
-        render_window->ForwardKeyPressEvent(event);
-    }
 }
-void GMainWindow::keyReleaseEvent(QKeyEvent* event) {
+void GMainWindow::AcceptDropEvent(QDropEvent* event) {
-    if (render_window) {
+    if (IsSingleFileDropEvent(event->mimeData())) {
-        render_window->ForwardKeyReleaseEvent(event);
+        event->setDropAction(Qt::DropAction::LinkAction);
+        event->accept();
    }
 }
-static bool IsSingleFileDropEvent(QDropEvent* event) {
+bool GMainWindow::DropAction(QDropEvent* event) {
-    const QMimeData* mimeData = event->mimeData();
+    if (!IsSingleFileDropEvent(event->mimeData())) {
-    return mimeData->hasUrls() && mimeData->urls().length() == 1;
+        return false;
-}
-void GMainWindow::dropEvent(QDropEvent* event) {
-    if (!IsSingleFileDropEvent(event)) {
-        return;
    }
    const QMimeData* mime_data = event->mimeData();
-    const QString filename = mime_data->urls().at(0).toLocalFile();
+    const QString& filename = mime_data->urls().at(0).toLocalFile();
    if (emulation_running && QFileInfo(filename).suffix() == QStringLiteral("bin")) {
+        // Amiibo
        LoadAmiibo(filename);
    } else {
+        // Game
        if (ConfirmChangeGame()) {
            BootGame(filename);
        }
    }
+    return true;
+}
+void GMainWindow::dropEvent(QDropEvent* event) {
+    DropAction(event);
 }
 void GMainWindow::dragEnterEvent(QDragEnterEvent* event) {
-    if (IsSingleFileDropEvent(event)) {
+    AcceptDropEvent(event);
-        event->acceptProposedAction();
-    }
 }
 void GMainWindow::dragMoveEvent(QDragMoveEvent* event) {
-    event->acceptProposedAction();
+    AcceptDropEvent(event);
 }
 bool GMainWindow::ConfirmChangeGame() {
@@ -2377,6 +2375,7 @@ int main(int argc, char* argv[]) {
    // Enables the core to make the qt created contexts current on std::threads
    QCoreApplication::setAttribute(Qt::AA_DontCheckOpenGLContextThreadAffinity);
+    QCoreApplication::setAttribute(Qt::AA_ShareOpenGLContexts);
    QApplication app(argc, argv);
    // Qt changes the locale and causes issues in float conversion using std::to_string() when
diff --git a/src/yuzu/main.h b/src/yuzu/main.h
index 8eba2172c..a67125567 100644
--- a/src/yuzu/main.h
+++ b/src/yuzu/main.h
@@ -78,6 +78,9 @@ public:
    std::unique_ptr<DiscordRPC::DiscordInterface> discord_rpc;
+    bool DropAction(QDropEvent* event);
+    void AcceptDropEvent(QDropEvent* event);
 signals:
    /**
@@ -264,8 +267,4 @@ protected:
    void dropEvent(QDropEvent* event) override;
    void dragEnterEvent(QDragEnterEvent* event) override;
    void dragMoveEvent(QDragMoveEvent* event) override;
-    // Overrides used to forward signals to the render window when the focus moves out.
-    void keyPressEvent(QKeyEvent* event) override;
-    void keyReleaseEvent(QKeyEvent* event) override;
 };
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index 96f1ce3af..b77c12baf 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -390,6 +390,8 @@ void Config::ReadValues() {
        sdl2_config->GetBoolean("Renderer", "use_accurate_gpu_emulation", false);
    Settings::values.use_asynchronous_gpu_emulation =
        sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", false);
+    Settings::values.use_vsync =
+        static_cast<u16>(sdl2_config->GetInteger("Renderer", "use_vsync", 1));
    Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
    Settings::values.bg_green =
diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h
index 8a2b658cd..085ffbc81 100644
--- a/src/yuzu_cmd/default_ini.h
+++ b/src/yuzu_cmd/default_ini.h
@@ -84,7 +84,7 @@ touch_device=
 # from any cemuhook compatible motion program.
 # IPv4 address of the udp input server (Default "127.0.0.1")
-udp_input_address=
+udp_input_address=127.0.0.1
 # Port of the udp input server. (Default 26760)
 udp_input_port=
@@ -150,6 +150,11 @@ use_accurate_gpu_emulation =
 # 0 : Off (slow), 1 (default): On (fast)
 use_asynchronous_gpu_emulation =
+# Forces VSync on the display thread. Usually doesn't impact performance, but on some drivers it can
+# so only turn this off if you notice a speed difference.
+# 0: Off, 1 (default): On
+use_vsync =
 # The clear color for the renderer. What shows up on the sides of the bottom screen.
 # Must be in range of 0.0-1.0. Defaults to 1.0 for all.
 bg_red =
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
index e96139885..19584360c 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
@@ -13,7 +13,7 @@
 #include "input_common/sdl/sdl.h"
 #include "yuzu_cmd/emu_window/emu_window_sdl2.h"
-EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
+EmuWindow_SDL2::EmuWindow_SDL2(Core::System& system, bool fullscreen) : system{system} {
    if (SDL_Init(SDL_INIT_VIDEO | SDL_INIT_JOYSTICK) < 0) {
        LOG_CRITICAL(Frontend, "Failed to initialize SDL2! Exiting...");
        exit(1);
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.h b/src/yuzu_cmd/emu_window/emu_window_sdl2.h
index b38f56661..fffac4252 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2.h
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.h
@@ -10,9 +10,13 @@
 struct SDL_Window;
+namespace Core {
+class System;
+}
 class EmuWindow_SDL2 : public Core::Frontend::EmuWindow {
 public:
-    explicit EmuWindow_SDL2(bool fullscreen);
+    explicit EmuWindow_SDL2(Core::System& system, bool fullscreen);
    ~EmuWindow_SDL2();
    /// Polls window events
@@ -24,6 +28,9 @@ public:
    /// Returns if window is shown (not minimized)
    bool IsShown() const override;
+    /// Presents the next frame
+    virtual void Present() = 0;
 protected:
    /// Called by PollEvents when a key is pressed or released.
    void OnKeyEvent(int key, u8 state);
@@ -55,6 +62,9 @@ protected:
    /// Called when a configuration change affects the minimal size of the window
    void OnMinimalClientAreaChangeRequest(std::pair<unsigned, unsigned> minimal_size) override;
+    /// Instance of the system, used to access renderer for the presentation thread
+    Core::System& system;
    /// Is the window still open?
    bool is_open = true;
@@ -62,7 +72,7 @@ protected:
    bool is_shown = true;
    /// Internal SDL2 render window
-    SDL_Window* render_window;
+    SDL_Window* render_window{};
    /// Keeps track of how often to update the title bar during gameplay
    u32 last_time = 0;
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp
index 7ffa0ac09..c0d373477 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.cpp
@@ -13,24 +13,25 @@
 #include "common/logging/log.h"
 #include "common/scm_rev.h"
 #include "common/string_util.h"
+#include "core/core.h"
 #include "core/settings.h"
 #include "input_common/keyboard.h"
 #include "input_common/main.h"
 #include "input_common/motion_emu.h"
+#include "video_core/renderer_base.h"
 #include "yuzu_cmd/emu_window/emu_window_sdl2_gl.h"
 class SDLGLContext : public Core::Frontend::GraphicsContext {
 public:
    explicit SDLGLContext() {
        // create a hidden window to make the shared context against
-        window = SDL_CreateWindow("", SDL_WINDOWPOS_UNDEFINED, // x position
+        window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
-                                  SDL_WINDOWPOS_UNDEFINED,     // y position
+                                  SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
-                                  Layout::ScreenUndocked::Width, Layout::ScreenUndocked::Height,
-                                  SDL_WINDOW_OPENGL | SDL_WINDOW_HIDDEN);
        context = SDL_GL_CreateContext(window);
    }
    ~SDLGLContext() {
+        DoneCurrent();
        SDL_GL_DeleteContext(context);
        SDL_DestroyWindow(window);
    }
@@ -43,8 +44,6 @@ public:
        SDL_GL_MakeCurrent(window, nullptr);
    }
-    void SwapBuffers() override {}
 private:
    SDL_Window* window;
    SDL_GLContext context;
@@ -80,7 +79,8 @@ bool EmuWindow_SDL2_GL::SupportsRequiredGLExtensions() {
    return unsupported_ext.empty();
 }
-EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscreen) {
+EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(Core::System& system, bool fullscreen)
+    : EmuWindow_SDL2{system, fullscreen} {
    SDL_GL_SetAttribute(SDL_GL_CONTEXT_MAJOR_VERSION, 4);
    SDL_GL_SetAttribute(SDL_GL_CONTEXT_MINOR_VERSION, 3);
    SDL_GL_SetAttribute(SDL_GL_CONTEXT_PROFILE_MASK, SDL_GL_CONTEXT_PROFILE_COMPATIBILITY);
@@ -90,6 +90,7 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscree
    SDL_GL_SetAttribute(SDL_GL_BLUE_SIZE, 8);
    SDL_GL_SetAttribute(SDL_GL_ALPHA_SIZE, 0);
    SDL_GL_SetAttribute(SDL_GL_SHARE_WITH_CURRENT_CONTEXT, 1);
+    SDL_GL_SetSwapInterval(0);
    std::string window_title = fmt::format("yuzu {} | {}-{}", Common::g_build_fullname,
                                           Common::g_scm_branch, Common::g_scm_desc);
@@ -105,13 +106,22 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscree
        exit(1);
    }
+    dummy_window = SDL_CreateWindow(NULL, SDL_WINDOWPOS_UNDEFINED, SDL_WINDOWPOS_UNDEFINED, 0, 0,
+                                    SDL_WINDOW_HIDDEN | SDL_WINDOW_OPENGL);
    if (fullscreen) {
        Fullscreen();
    }
-    gl_context = SDL_GL_CreateContext(render_window);
-    if (gl_context == nullptr) {
+    window_context = SDL_GL_CreateContext(render_window);
-        LOG_CRITICAL(Frontend, "Failed to create SDL2 GL context! {}", SDL_GetError());
+    core_context = CreateSharedContext();
+    if (window_context == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create SDL2 GL context: {}", SDL_GetError());
+        exit(1);
+    }
+    if (core_context == nullptr) {
+        LOG_CRITICAL(Frontend, "Failed to create shared SDL2 GL context: {}", SDL_GetError());
        exit(1);
    }
@@ -128,28 +138,22 @@ EmuWindow_SDL2_GL::EmuWindow_SDL2_GL(bool fullscreen) : EmuWindow_SDL2(fullscree
    OnResize();
    OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
    SDL_PumpEvents();
-    SDL_GL_SetSwapInterval(false);
    LOG_INFO(Frontend, "yuzu Version: {} | {}-{}", Common::g_build_fullname, Common::g_scm_branch,
             Common::g_scm_desc);
    Settings::LogSettings();
-    DoneCurrent();
 }
 EmuWindow_SDL2_GL::~EmuWindow_SDL2_GL() {
-    SDL_GL_DeleteContext(gl_context);
+    core_context.reset();
-}
+    SDL_GL_DeleteContext(window_context);
-void EmuWindow_SDL2_GL::SwapBuffers() {
-    SDL_GL_SwapWindow(render_window);
 }
 void EmuWindow_SDL2_GL::MakeCurrent() {
-    SDL_GL_MakeCurrent(render_window, gl_context);
+    core_context->MakeCurrent();
 }
 void EmuWindow_SDL2_GL::DoneCurrent() {
-    SDL_GL_MakeCurrent(render_window, nullptr);
+    core_context->DoneCurrent();
 }
 void EmuWindow_SDL2_GL::RetrieveVulkanHandlers(void* get_instance_proc_addr, void* instance,
@@ -161,3 +165,13 @@ void EmuWindow_SDL2_GL::RetrieveVulkanHandlers(void* get_instance_proc_addr, voi
 std::unique_ptr<Core::Frontend::GraphicsContext> EmuWindow_SDL2_GL::CreateSharedContext() const {
    return std::make_unique<SDLGLContext>();
 }
+void EmuWindow_SDL2_GL::Present() {
+    SDL_GL_MakeCurrent(render_window, window_context);
+    SDL_GL_SetSwapInterval(Settings::values.use_vsync ? 1 : 0);
+    while (IsOpen()) {
+        system.Renderer().TryPresent(100);
+        SDL_GL_SwapWindow(render_window);
+    }
+    SDL_GL_MakeCurrent(render_window, nullptr);
+}
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h
index c753085a8..b80669ff0 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_gl.h
@@ -10,17 +10,12 @@
 class EmuWindow_SDL2_GL final : public EmuWindow_SDL2 {
 public:
-    explicit EmuWindow_SDL2_GL(bool fullscreen);
+    explicit EmuWindow_SDL2_GL(Core::System& system, bool fullscreen);
    ~EmuWindow_SDL2_GL();
-    /// Swap buffers to display the next frame
-    void SwapBuffers() override;
-    /// Makes the graphics context current for the caller thread
    void MakeCurrent() override;
-    /// Releases the GL context from the caller thread
    void DoneCurrent() override;
+    void Present() override;
    /// Ignored in OpenGL
    void RetrieveVulkanHandlers(void* get_instance_proc_addr, void* instance,
@@ -29,10 +24,17 @@ public:
    std::unique_ptr<Core::Frontend::GraphicsContext> CreateSharedContext() const override;
 private:
+    /// Fake hidden window for the core context
+    SDL_Window* dummy_window{};
    /// Whether the GPU and driver supports the OpenGL extension required
    bool SupportsRequiredGLExtensions();
    using SDL_GLContext = void*;
    /// The OpenGL context associated with the window
-    SDL_GLContext gl_context;
+    SDL_GLContext window_context;
+    /// The OpenGL context associated with the core
+    std::unique_ptr<Core::Frontend::GraphicsContext> core_context;
 };
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
index a203f0da9..abcc58165 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
@@ -15,7 +15,8 @@
 #include "core/settings.h"
 #include "yuzu_cmd/emu_window/emu_window_sdl2_vk.h"
-EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(bool fullscreen) : EmuWindow_SDL2(fullscreen) {
+EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(Core::System& system, bool fullscreen)
+    : EmuWindow_SDL2{system, fullscreen} {
    if (SDL_Vulkan_LoadLibrary(nullptr) != 0) {
        LOG_CRITICAL(Frontend, "SDL failed to load the Vulkan library: {}", SDL_GetError());
        exit(EXIT_FAILURE);
@@ -110,8 +111,6 @@ EmuWindow_SDL2_VK::~EmuWindow_SDL2_VK() {
    vkDestroyInstance(vk_instance, nullptr);
 }
-void EmuWindow_SDL2_VK::SwapBuffers() {}
 void EmuWindow_SDL2_VK::MakeCurrent() {
    // Unused on Vulkan
 }
@@ -160,3 +159,7 @@ bool EmuWindow_SDL2_VK::UseStandardLayers(PFN_vkGetInstanceProcAddr vkGetInstanc
               return layer.layerName == std::string("VK_LAYER_LUNARG_standard_validation");
           }) != layers.end();
 }
+void EmuWindow_SDL2_VK::Present() {
+    // TODO (bunnei): ImplementMe
+}
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
index 2a7c06a24..1eb8c0868 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
@@ -10,19 +10,12 @@
 class EmuWindow_SDL2_VK final : public EmuWindow_SDL2 {
 public:
-    explicit EmuWindow_SDL2_VK(bool fullscreen);
+    explicit EmuWindow_SDL2_VK(Core::System& system, bool fullscreen);
    ~EmuWindow_SDL2_VK();
-    /// Swap buffers to display the next frame
-    void SwapBuffers() override;
-    /// Makes the graphics context current for the caller thread
    void MakeCurrent() override;
-    /// Releases the GL context from the caller thread
    void DoneCurrent() override;
+    void Present() override;
-    /// Retrieves Vulkan specific handlers from the window
    void RetrieveVulkanHandlers(void* get_instance_proc_addr, void* instance,
                                void* surface) const override;
diff --git a/src/yuzu_cmd/yuzu.cpp b/src/yuzu_cmd/yuzu.cpp
index 325795321..babf4c3a4 100644
--- a/src/yuzu_cmd/yuzu.cpp
+++ b/src/yuzu_cmd/yuzu.cpp
@@ -177,14 +177,16 @@ int main(int argc, char** argv) {
    Settings::values.use_gdbstub = use_gdbstub;
    Settings::Apply();
+    Core::System& system{Core::System::GetInstance()};
    std::unique_ptr<EmuWindow_SDL2> emu_window;
    switch (Settings::values.renderer_backend) {
    case Settings::RendererBackend::OpenGL:
-        emu_window = std::make_unique<EmuWindow_SDL2_GL>(fullscreen);
+        emu_window = std::make_unique<EmuWindow_SDL2_GL>(system, fullscreen);
        break;
    case Settings::RendererBackend::Vulkan:
 #ifdef HAS_VULKAN
-        emu_window = std::make_unique<EmuWindow_SDL2_VK>(fullscreen);
+        emu_window = std::make_unique<EmuWindow_SDL2_VK>(system, fullscreen);
        break;
 #else
        LOG_CRITICAL(Frontend, "Vulkan backend has not been compiled!");
@@ -192,12 +194,6 @@ int main(int argc, char** argv) {
 #endif
    }
-    if (!Settings::values.use_multi_core) {
-        // Single core mode must acquire OpenGL context for entire emulation session
-        emu_window->MakeCurrent();
-    }
-    Core::System& system{Core::System::GetInstance()};
    system.SetContentProvider(std::make_unique<FileSys::ContentProviderUnion>());
    system.SetFilesystem(std::make_shared<FileSys::RealVfsFilesystem>());
    system.GetFileSystemController().CreateFactories(*system.GetFilesystem());
@@ -234,12 +230,23 @@ int main(int argc, char** argv) {
    system.TelemetrySession().AddField(Telemetry::FieldType::App, "Frontend", "SDL");
-    emu_window->MakeCurrent();
    system.Renderer().Rasterizer().LoadDiskResources();
+    // Acquire render context for duration of the thread if this is the rendering thread
+    if (!Settings::values.use_asynchronous_gpu_emulation) {
+        emu_window->MakeCurrent();
+    }
+    SCOPE_EXIT({
+        if (!Settings::values.use_asynchronous_gpu_emulation) {
+            emu_window->DoneCurrent();
+        }
+    });
+    std::thread render_thread([&emu_window] { emu_window->Present(); });
    while (emu_window->IsOpen()) {
        system.RunLoop();
    }
+    render_thread.join();
    system.Shutdown();
diff --git a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp
index f2cc4a797..a1bdb1a12 100644
--- a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp
+++ b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.cpp
@@ -112,10 +112,6 @@ EmuWindow_SDL2_Hide::~EmuWindow_SDL2_Hide() {
    SDL_Quit();
 }
-void EmuWindow_SDL2_Hide::SwapBuffers() {
-    SDL_GL_SwapWindow(render_window);
-}
 void EmuWindow_SDL2_Hide::PollEvents() {}
 void EmuWindow_SDL2_Hide::MakeCurrent() {
diff --git a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h
index c7fccc002..b13e15309 100644
--- a/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h
+++ b/src/yuzu_tester/emu_window/emu_window_sdl2_hide.h
@@ -13,9 +13,6 @@ public:
    explicit EmuWindow_SDL2_Hide();
    ~EmuWindow_SDL2_Hide();
-    /// Swap buffers to display the next frame
-    void SwapBuffers() override;
    /// Polls window events
    void PollEvents() override;