107 files changed, 1079 insertions, 1602 deletions
diff --git a/.ci/templates/build-msvc.yml b/.ci/templates/build-msvc.yml
index 33ff8201f..721179550 100644
--- a/.ci/templates/build-msvc.yml
+++ b/.ci/templates/build-msvc.yml
@@ -8,7 +8,7 @@ steps:
  displayName: 'Install vulkan-sdk'
 - script: python -m pip install --upgrade pip conan
  displayName: 'Install conan'
- script: refreshenv && mkdir build && cd build && cmake -G "Visual Studio 16 2019" -A x64 --config Release -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_QT_WEB_ENGINE=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON -DDISPLAY_VERSION=${{ parameters['version'] }} .. && cd ..
+- script: refreshenv && mkdir build && cd build && cmake -G "Visual Studio 16 2019" -A x64 --config Release -DYUZU_USE_BUNDLED_QT=1 -DYUZU_USE_QT_WEB_ENGINE=ON -DENABLE_COMPATIBILITY_LIST_DOWNLOAD=ON -DYUZU_ENABLE_COMPATIBILITY_REPORTING=${COMPAT} -DUSE_DISCORD_PRESENCE=ON -DENABLE_QT_TRANSLATION=ON -DDISPLAY_VERSION=${{ parameters['version'] }} .. && cd ..
  displayName: 'Configure CMake'
 - task: MSBuild@1
  displayName: 'Build'
diff --git a/CMakeLists.txt b/CMakeLists.txt
index eda555494..5793c5d57 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -165,7 +165,7 @@ macro(yuzu_find_packages)
        "lz4               1.8         lz4/1.9.2"
        "nlohmann_json     3.8         nlohmann_json/3.8.0"
        "ZLIB              1.2         zlib/1.2.11"
-        "zstd              1.4         zstd/1.4.5"
+        "zstd              1.4         zstd/1.4.8"
    )
    foreach(PACKAGE ${REQUIRED_LIBS})
diff --git a/externals/dynarmic b/externals/dynarmic
-Subproject 4a9a0d07f7376b439430e686721e8176c7b56ce
+Subproject 3806284cbefc4115436dcdc687776a45ec31309
diff --git a/src/audio_core/algorithm/interpolate.cpp b/src/audio_core/algorithm/interpolate.cpp
index 699fcb84c..3b4144e21 100644
--- a/src/audio_core/algorithm/interpolate.cpp
+++ b/src/audio_core/algorithm/interpolate.cpp
@@ -218,7 +218,7 @@ void Resample(s32* output, const s32* input, s32 pitch, s32& fraction, std::size
        const auto l2 = lut[lut_index + 2];
        const auto l3 = lut[lut_index + 3];
-        const auto s0 = static_cast<s32>(input[index]);
+        const auto s0 = static_cast<s32>(input[index + 0]);
        const auto s1 = static_cast<s32>(input[index + 1]);
        const auto s2 = static_cast<s32>(input[index + 2]);
        const auto s3 = static_cast<s32>(input[index + 3]);
diff --git a/src/audio_core/audio_renderer.cpp b/src/audio_core/audio_renderer.cpp
index 179560cd7..d2ce8c814 100644
--- a/src/audio_core/audio_renderer.cpp
+++ b/src/audio_core/audio_renderer.cpp
@@ -11,7 +11,6 @@
 #include "audio_core/info_updater.h"
 #include "audio_core/voice_context.h"
 #include "common/logging/log.h"
-#include "core/hle/kernel/writable_event.h"
 #include "core/memory.h"
 #include "core/settings.h"
@@ -71,10 +70,9 @@ namespace {
 namespace AudioCore {
 AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory::Memory& memory_,
                             AudioCommon::AudioRendererParameter params,
-                             std::shared_ptr<Kernel::WritableEvent> buffer_event_,
+                             Stream::ReleaseCallback&& release_callback,
                             std::size_t instance_number)
-    : worker_params{params}, buffer_event{buffer_event_},
+    : worker_params{params}, memory_pool_info(params.effect_count + params.voice_count * 4),
-      memory_pool_info(params.effect_count + params.voice_count * 4),
      voice_context(params.voice_count), effect_context(params.effect_count), mix_context(),
      sink_context(params.sink_count), splitter_context(),
      voices(params.voice_count), memory{memory_},
@@ -85,10 +83,9 @@ AudioRenderer::AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory
                                params.num_splitter_send_channels);
    mix_context.Initialize(behavior_info, params.submix_count + 1, params.effect_count);
    audio_out = std::make_unique<AudioCore::AudioOut>();
-    stream =
+    stream = audio_out->OpenStream(
-        audio_out->OpenStream(core_timing, params.sample_rate, AudioCommon::STREAM_NUM_CHANNELS,
+        core_timing, params.sample_rate, AudioCommon::STREAM_NUM_CHANNELS,
-                              fmt::format("AudioRenderer-Instance{}", instance_number),
+        fmt::format("AudioRenderer-Instance{}", instance_number), std::move(release_callback));
-                              [=]() { buffer_event_->Signal(); });
    audio_out->StartStream(stream);
    QueueMixedBuffer(0);
diff --git a/src/audio_core/audio_renderer.h b/src/audio_core/audio_renderer.h
index 90f7eafa4..18567f618 100644
--- a/src/audio_core/audio_renderer.h
+++ b/src/audio_core/audio_renderer.h
@@ -27,10 +27,6 @@ namespace Core::Timing {
 class CoreTiming;
 }
-namespace Kernel {
-class WritableEvent;
-}
 namespace Core::Memory {
 class Memory;
 }
@@ -44,8 +40,7 @@ class AudioRenderer {
 public:
    AudioRenderer(Core::Timing::CoreTiming& core_timing, Core::Memory::Memory& memory_,
                  AudioCommon::AudioRendererParameter params,
-                  std::shared_ptr<Kernel::WritableEvent> buffer_event_,
+                  Stream::ReleaseCallback&& release_callback, std::size_t instance_number);
-                  std::size_t instance_number);
    ~AudioRenderer();
    [[nodiscard]] ResultCode UpdateAudioRenderer(const std::vector<u8>& input_params,
@@ -61,7 +56,6 @@ private:
    BehaviorInfo behavior_info{};
    AudioCommon::AudioRendererParameter worker_params;
-    std::shared_ptr<Kernel::WritableEvent> buffer_event;
    std::vector<ServerMemoryPoolInfo> memory_pool_info;
    VoiceContext voice_context;
    EffectContext effect_context;
diff --git a/src/audio_core/stream.cpp b/src/audio_core/stream.cpp
index eca296589..afe68c9ed 100644
--- a/src/audio_core/stream.cpp
+++ b/src/audio_core/stream.cpp
@@ -130,7 +130,11 @@ bool Stream::ContainsBuffer([[maybe_unused]] Buffer::Tag tag) const {
 std::vector<Buffer::Tag> Stream::GetTagsAndReleaseBuffers(std::size_t max_count) {
    std::vector<Buffer::Tag> tags;
    for (std::size_t count = 0; count < max_count && !released_buffers.empty(); ++count) {
-        tags.push_back(released_buffers.front()->GetTag());
+        if (released_buffers.front()) {
+            tags.push_back(released_buffers.front()->GetTag());
+        } else {
+            ASSERT_MSG(false, "Invalid tag in released_buffers!");
+        }
        released_buffers.pop();
    }
    return tags;
@@ -140,7 +144,11 @@ std::vector<Buffer::Tag> Stream::GetTagsAndReleaseBuffers() {
    std::vector<Buffer::Tag> tags;
    tags.reserve(released_buffers.size());
    while (!released_buffers.empty()) {
-        tags.push_back(released_buffers.front()->GetTag());
+        if (released_buffers.front()) {
+            tags.push_back(released_buffers.front()->GetTag());
+        } else {
+            ASSERT_MSG(false, "Invalid tag in released_buffers!");
+        }
        released_buffers.pop();
    }
    return tags;
diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 943ff996e..2c2bd2ee8 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -135,8 +135,6 @@ add_library(common STATIC
    math_util.h
    memory_detect.cpp
    memory_detect.h
-    memory_hook.cpp
-    memory_hook.h
    microprofile.cpp
    microprofile.h
    microprofileui.h
@@ -162,6 +160,8 @@ add_library(common STATIC
    thread.cpp
    thread.h
    thread_queue_list.h
+    thread_worker.cpp
+    thread_worker.h
    threadsafe_queue.h
    time_zone.cpp
    time_zone.h
diff --git a/src/common/memory_hook.cpp b/src/common/memory_hook.cpp
deleted file mode 100644
index 3986986d6..000000000
--- a/src/common/memory_hook.cpp
+++ /dev/null
@@ -1,11 +0,0 @@
-// Copyright 2018 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#include "common/memory_hook.h"
-namespace Common {
-MemoryHook::~MemoryHook() = default;
-} // namespace Common
diff --git a/src/common/memory_hook.h b/src/common/memory_hook.h
deleted file mode 100644
index adaa4c2c5..000000000
--- a/src/common/memory_hook.h
+++ /dev/null
@@ -1,47 +0,0 @@
-// Copyright 2016 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#pragma once
-#include <memory>
-#include <optional>
-#include "common/common_types.h"
-namespace Common {
-/**
- * Memory hooks have two purposes:
- * 1. To allow reads and writes to a region of memory to be intercepted. This is used to implement
- *    texture forwarding and memory breakpoints for debugging.
- * 2. To allow for the implementation of MMIO devices.
- *
- * A hook may be mapped to multiple regions of memory.
- *
- * If a std::nullopt or false is returned from a function, the read/write request is passed through
- * to the underlying memory region.
- */
-class MemoryHook {
-public:
-    virtual ~MemoryHook();
-    virtual std::optional<bool> IsValidAddress(VAddr addr) = 0;
-    virtual std::optional<u8> Read8(VAddr addr) = 0;
-    virtual std::optional<u16> Read16(VAddr addr) = 0;
-    virtual std::optional<u32> Read32(VAddr addr) = 0;
-    virtual std::optional<u64> Read64(VAddr addr) = 0;
-    virtual bool ReadBlock(VAddr src_addr, void* dest_buffer, std::size_t size) = 0;
-    virtual bool Write8(VAddr addr, u8 data) = 0;
-    virtual bool Write16(VAddr addr, u16 data) = 0;
-    virtual bool Write32(VAddr addr, u32 data) = 0;
-    virtual bool Write64(VAddr addr, u64 data) = 0;
-    virtual bool WriteBlock(VAddr dest_addr, const void* src_buffer, std::size_t size) = 0;
-};
-using MemoryHookPointer = std::shared_ptr<MemoryHook>;
-} // namespace Common
diff --git a/src/common/page_table.cpp b/src/common/page_table.cpp
index bccea0894..8fd8620fd 100644
--- a/src/common/page_table.cpp
+++ b/src/common/page_table.cpp
@@ -10,16 +10,10 @@ PageTable::PageTable() = default;
 PageTable::~PageTable() noexcept = default;
-void PageTable::Resize(std::size_t address_space_width_in_bits, std::size_t page_size_in_bits,
+void PageTable::Resize(size_t address_space_width_in_bits, size_t page_size_in_bits) {
-                       bool has_attribute) {
+    const size_t num_page_table_entries{1ULL << (address_space_width_in_bits - page_size_in_bits)};
-    const std::size_t num_page_table_entries{1ULL
-                                             << (address_space_width_in_bits - page_size_in_bits)};
    pointers.resize(num_page_table_entries);
    backing_addr.resize(num_page_table_entries);
-    if (has_attribute) {
-        attributes.resize(num_page_table_entries);
-    }
 }
 } // namespace Common
diff --git a/src/common/page_table.h b/src/common/page_table.h
index 9754fabf9..61c5552e0 100644
--- a/src/common/page_table.h
+++ b/src/common/page_table.h
@@ -4,10 +4,10 @@
 #pragma once
+#include <atomic>
 #include <tuple>
 #include "common/common_types.h"
-#include "common/memory_hook.h"
 #include "common/virtual_buffer.h"
 namespace Common {
@@ -20,27 +20,6 @@ enum class PageType : u8 {
    /// Page is mapped to regular memory, but also needs to check for rasterizer cache flushing and
    /// invalidation
    RasterizerCachedMemory,
-    /// Page is mapped to a I/O region. Writing and reading to this page is handled by functions.
-    Special,
-    /// Page is allocated for use.
-    Allocated,
-};
-struct SpecialRegion {
-    enum class Type {
-        DebugHook,
-        IODevice,
-    } type;
-    MemoryHookPointer handler;
-    [[nodiscard]] bool operator<(const SpecialRegion& other) const {
-        return std::tie(type, handler) < std::tie(other.type, other.handler);
-    }
-    [[nodiscard]] bool operator==(const SpecialRegion& other) const {
-        return std::tie(type, handler) == std::tie(other.type, other.handler);
-    }
 };
 /**
@@ -48,6 +27,59 @@ struct SpecialRegion {
 * mimics the way a real CPU page table works.
 */
 struct PageTable {
+    /// Number of bits reserved for attribute tagging.
+    /// This can be at most the guaranteed alignment of the pointers in the page table.
+    static constexpr int ATTRIBUTE_BITS = 2;
+    /**
+     * Pair of host pointer and page type attribute.
+     * This uses the lower bits of a given pointer to store the attribute tag.
+     * Writing and reading the pointer attribute pair is guaranteed to be atomic for the same method
+     * call. In other words, they are guaranteed to be synchronized at all times.
+     */
+    class PageInfo {
+    public:
+        /// Returns the page pointer
+        [[nodiscard]] u8* Pointer() const noexcept {
+            return ExtractPointer(raw.load(std::memory_order_relaxed));
+        }
+        /// Returns the page type attribute
+        [[nodiscard]] PageType Type() const noexcept {
+            return ExtractType(raw.load(std::memory_order_relaxed));
+        }
+        /// Returns the page pointer and attribute pair, extracted from the same atomic read
+        [[nodiscard]] std::pair<u8*, PageType> PointerType() const noexcept {
+            const uintptr_t non_atomic_raw = raw.load(std::memory_order_relaxed);
+            return {ExtractPointer(non_atomic_raw), ExtractType(non_atomic_raw)};
+        }
+        /// Returns the raw representation of the page information.
+        /// Use ExtractPointer and ExtractType to unpack the value.
+        [[nodiscard]] uintptr_t Raw() const noexcept {
+            return raw.load(std::memory_order_relaxed);
+        }
+        /// Write a page pointer and type pair atomically
+        void Store(u8* pointer, PageType type) noexcept {
+            raw.store(reinterpret_cast<uintptr_t>(pointer) | static_cast<uintptr_t>(type));
+        }
+        /// Unpack a pointer from a page info raw representation
+        [[nodiscard]] static u8* ExtractPointer(uintptr_t raw) noexcept {
+            return reinterpret_cast<u8*>(raw & (~uintptr_t{0} << ATTRIBUTE_BITS));
+        }
+        /// Unpack a page type from a page info raw representation
+        [[nodiscard]] static PageType ExtractType(uintptr_t raw) noexcept {
+            return static_cast<PageType>(raw & ((uintptr_t{1} << ATTRIBUTE_BITS) - 1));
+        }
+    private:
+        std::atomic<uintptr_t> raw;
+    };
    PageTable();
    ~PageTable() noexcept;
@@ -58,25 +90,21 @@ struct PageTable {
    PageTable& operator=(PageTable&&) noexcept = default;
    /**
-     * Resizes the page table to be able to accomodate enough pages within
+     * Resizes the page table to be able to accommodate enough pages within
     * a given address space.
     *
     * @param address_space_width_in_bits The address size width in bits.
     * @param page_size_in_bits           The page size in bits.
-     * @param has_attribute               Whether or not this page has any backing attributes.
     */
-    void Resize(std::size_t address_space_width_in_bits, std::size_t page_size_in_bits,
+    void Resize(size_t address_space_width_in_bits, size_t page_size_in_bits);
-                bool has_attribute);
    /**
     * Vector of memory pointers backing each page. An entry can only be non-null if the
-     * corresponding entry in the `attributes` vector is of type `Memory`.
+     * corresponding attribute element is of type `Memory`.
     */
-    VirtualBuffer<u8*> pointers;
+    VirtualBuffer<PageInfo> pointers;
    VirtualBuffer<u64> backing_addr;
-    VirtualBuffer<PageType> attributes;
 };
 } // namespace Common
diff --git a/src/common/swap.h b/src/common/swap.h
index 7665942a2..a80e191dc 100644
--- a/src/common/swap.h
+++ b/src/common/swap.h
@@ -394,7 +394,7 @@ public:
    template <typename S, typename T2, typename F2>
    friend S operator%(const S& p, const swapped_t v);
-    // Arithmetics + assignements
+    // Arithmetics + assignments
    template <typename S, typename T2, typename F2>
    friend S operator+=(const S& p, const swapped_t v);
@@ -451,7 +451,7 @@ S operator%(const S& i, const swap_struct_t<T, F> v) {
    return i % v.swap();
 }
-// Arithmetics + assignements
+// Arithmetics + assignments
 template <typename S, typename T, typename F>
 S& operator+=(S& i, const swap_struct_t<T, F> v) {
    i += v.swap();
diff --git a/src/common/thread_worker.cpp b/src/common/thread_worker.cpp
new file mode 100644
index 000000000..8f9bf447a
--- /dev/null
+++ b/src/common/thread_worker.cpp
@@ -0,0 +1,58 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#include "common/thread.h"
+#include "common/thread_worker.h"
+namespace Common {
+ThreadWorker::ThreadWorker(std::size_t num_workers, const std::string& name) {
+    for (std::size_t i = 0; i < num_workers; ++i)
+        threads.emplace_back([this, thread_name{std::string{name}}] {
+            Common::SetCurrentThreadName(thread_name.c_str());
+            // Wait for first request
+            {
+                std::unique_lock lock{queue_mutex};
+                condition.wait(lock, [this] { return stop || !requests.empty(); });
+            }
+            while (true) {
+                std::function<void()> task;
+                {
+                    std::unique_lock lock{queue_mutex};
+                    condition.wait(lock, [this] { return stop || !requests.empty(); });
+                    if (stop || requests.empty()) {
+                        return;
+                    }
+                    task = std::move(requests.front());
+                    requests.pop();
+                }
+                task();
+            }
+        });
+}
+ThreadWorker::~ThreadWorker() {
+    {
+        std::unique_lock lock{queue_mutex};
+        stop = true;
+    }
+    condition.notify_all();
+    for (std::thread& thread : threads) {
+        thread.join();
+    }
+}
+void ThreadWorker::QueueWork(std::function<void()>&& work) {
+    {
+        std::unique_lock lock{queue_mutex};
+        requests.emplace(work);
+    }
+    condition.notify_one();
+}
+} // namespace Common
diff --git a/src/common/thread_worker.h b/src/common/thread_worker.h
new file mode 100644
index 000000000..f1859971f
--- /dev/null
+++ b/src/common/thread_worker.h
@@ -0,0 +1,30 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#pragma once
+#include <atomic>
+#include <functional>
+#include <mutex>
+#include <string>
+#include <vector>
+#include <queue>
+namespace Common {
+class ThreadWorker final {
+public:
+    explicit ThreadWorker(std::size_t num_workers, const std::string& name);
+    ~ThreadWorker();
+    void QueueWork(std::function<void()>&& work);
+private:
+    std::vector<std::thread> threads;
+    std::queue<std::function<void()>> requests;
+    std::mutex queue_mutex;
+    std::condition_variable condition;
+    std::atomic_bool stop{};
+};
+} // namespace Common
diff --git a/src/common/virtual_buffer.h b/src/common/virtual_buffer.h
index 91d430036..fb1a6f81f 100644
--- a/src/common/virtual_buffer.h
+++ b/src/common/virtual_buffer.h
@@ -15,10 +15,12 @@ void FreeMemoryPages(void* base, std::size_t size) noexcept;
 template <typename T>
 class VirtualBuffer final {
 public:
-    static_assert(
+    // TODO: Uncomment this and change Common::PageTable::PageInfo to be trivially constructible
-        std::is_trivially_constructible_v<T>,
+    // using std::atomic_ref once libc++ has support for it
-        "T must be trivially constructible, as non-trivial constructors will not be executed "
+    // static_assert(
-        "with the current allocator");
+    //     std::is_trivially_constructible_v<T>,
+    //     "T must be trivially constructible, as non-trivial constructors will not be executed "
+    //     "with the current allocator");
    constexpr VirtualBuffer() = default;
    explicit VirtualBuffer(std::size_t count) : alloc_size{count * sizeof(T)} {
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index 59bd3d2a6..01f3e9419 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -202,6 +202,8 @@ add_library(core STATIC
    hle/kernel/server_port.h
    hle/kernel/server_session.cpp
    hle/kernel/server_session.h
+    hle/kernel/service_thread.cpp
+    hle/kernel/service_thread.h
    hle/kernel/session.cpp
    hle/kernel/session.h
    hle/kernel/shared_memory.cpp
@@ -500,7 +502,6 @@ add_library(core STATIC
    hle/service/sm/controller.h
    hle/service/sm/sm.cpp
    hle/service/sm/sm.h
-    hle/service/sockets/blocking_worker.h
    hle/service/sockets/bsd.cpp
    hle/service/sockets/bsd.h
    hle/service/sockets/ethc.cpp
diff --git a/src/core/arm/dynarmic/arm_dynarmic_32.cpp b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
index e9c74b1a6..0831dd5d2 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_32.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic_32.cpp
@@ -133,6 +133,7 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable&
    config.page_table = reinterpret_cast<std::array<std::uint8_t*, NUM_PAGE_TABLE_ENTRIES>*>(
        page_table.pointers.data());
    config.absolute_offset_page_table = true;
+    config.page_table_pointer_mask_bits = Common::PageTable::ATTRIBUTE_BITS;
    config.detect_misaligned_access_via_page_table = 16 | 32 | 64 | 128;
    config.only_detect_misalignment_via_page_table_on_page_boundary = true;
@@ -180,6 +181,9 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable&
        if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
        }
+        if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
+            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
+        }
    }
    return std::make_unique<Dynarmic::A32::Jit>(config);
diff --git a/src/core/arm/dynarmic/arm_dynarmic_64.cpp b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
index 7a4eb88a2..4c5ebca22 100644
--- a/src/core/arm/dynarmic/arm_dynarmic_64.cpp
+++ b/src/core/arm/dynarmic/arm_dynarmic_64.cpp
@@ -152,6 +152,7 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable&
    // Memory
    config.page_table = reinterpret_cast<void**>(page_table.pointers.data());
    config.page_table_address_space_bits = address_space_bits;
+    config.page_table_pointer_mask_bits = Common::PageTable::ATTRIBUTE_BITS;
    config.silently_mirror_page_table = false;
    config.absolute_offset_page_table = true;
    config.detect_misaligned_access_via_page_table = 16 | 32 | 64 | 128;
@@ -211,6 +212,9 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable&
        if (Settings::values.cpuopt_unsafe_reduce_fp_error) {
            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_ReducedErrorFP;
        }
+        if (Settings::values.cpuopt_unsafe_inaccurate_nan) {
+            config.optimizations |= Dynarmic::OptimizationFlag::Unsafe_InaccurateNaN;
+        }
    }
    return std::make_shared<Dynarmic::A64::Jit>(config);
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 0961c0819..1a2002dec 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -159,7 +159,7 @@ struct System::Impl {
        device_memory = std::make_unique<Core::DeviceMemory>();
        is_multicore = Settings::values.use_multi_core.GetValue();
-        is_async_gpu = is_multicore || Settings::values.use_asynchronous_gpu_emulation.GetValue();
+        is_async_gpu = Settings::values.use_asynchronous_gpu_emulation.GetValue();
        kernel.SetMulticore(is_multicore);
        cpu_manager.SetMulticore(is_multicore);
@@ -307,7 +307,6 @@ struct System::Impl {
        service_manager.reset();
        cheat_engine.reset();
        telemetry_session.reset();
-        device_memory.reset();
        // Close all CPU/threading state
        cpu_manager.Shutdown();
diff --git a/src/core/hle/kernel/hle_ipc.cpp b/src/core/hle/kernel/hle_ipc.cpp
index e75e80ad0..83decf6cf 100644
--- a/src/core/hle/kernel/hle_ipc.cpp
+++ b/src/core/hle/kernel/hle_ipc.cpp
@@ -46,43 +46,6 @@ void SessionRequestHandler::ClientDisconnected(
    boost::range::remove_erase(connected_sessions, server_session);
 }
-std::shared_ptr<WritableEvent> HLERequestContext::SleepClientThread(
-    const std::string& reason, u64 timeout, WakeupCallback&& callback,
-    std::shared_ptr<WritableEvent> writable_event) {
-    // Put the client thread to sleep until the wait event is signaled or the timeout expires.
-    if (!writable_event) {
-        // Create event if not provided
-        const auto pair = WritableEvent::CreateEventPair(kernel, "HLE Pause Event: " + reason);
-        writable_event = pair.writable;
-    }
-    Handle event_handle = InvalidHandle;
-    {
-        KScopedSchedulerLockAndSleep lock(kernel, event_handle, thread.get(), timeout);
-        thread->SetHLECallback(
-            [context = *this, callback](std::shared_ptr<Thread> thread) mutable -> bool {
-                ThreadWakeupReason reason = thread->GetSignalingResult() == RESULT_TIMEOUT
-                                                ? ThreadWakeupReason::Timeout
-                                                : ThreadWakeupReason::Signal;
-                callback(thread, context, reason);
-                context.WriteToOutgoingCommandBuffer(*thread);
-                return true;
-            });
-        const auto readable_event{writable_event->GetReadableEvent()};
-        writable_event->Clear();
-        thread->SetHLESyncObject(readable_event.get());
-        thread->SetStatus(ThreadStatus::WaitHLEEvent);
-        thread->SetSynchronizationResults(nullptr, RESULT_TIMEOUT);
-        readable_event->AddWaitingThread(thread);
-    }
-    thread->SetHLETimeEvent(event_handle);
-    is_thread_waiting = true;
-    return writable_event;
-}
 HLERequestContext::HLERequestContext(KernelCore& kernel, Core::Memory::Memory& memory,
                                     std::shared_ptr<ServerSession> server_session,
                                     std::shared_ptr<Thread> thread)
diff --git a/src/core/hle/kernel/hle_ipc.h b/src/core/hle/kernel/hle_ipc.h
index c31a65476..b112e1ebd 100644
--- a/src/core/hle/kernel/hle_ipc.h
+++ b/src/core/hle/kernel/hle_ipc.h
@@ -129,23 +129,6 @@ public:
    using WakeupCallback = std::function<void(
        std::shared_ptr<Thread> thread, HLERequestContext& context, ThreadWakeupReason reason)>;
-    /**
-     * Puts the specified guest thread to sleep until the returned event is signaled or until the
-     * specified timeout expires.
-     * @param reason Reason for pausing the thread, to be used for debugging purposes.
-     * @param timeout Timeout in nanoseconds after which the thread will be awoken and the callback
-     * invoked with a Timeout reason.
-     * @param callback Callback to be invoked when the thread is resumed. This callback must write
-     * the entire command response once again, regardless of the state of it before this function
-     * was called.
-     * @param writable_event Event to use to wake up the thread. If unspecified, an event will be
-     * created.
-     * @returns Event that when signaled will resume the thread and call the callback function.
-     */
-    std::shared_ptr<WritableEvent> SleepClientThread(
-        const std::string& reason, u64 timeout, WakeupCallback&& callback,
-        std::shared_ptr<WritableEvent> writable_event = nullptr);
    /// Populates this context with data from the requesting process/thread.
    ResultCode PopulateFromIncomingCommandBuffer(const HandleTable& handle_table,
                                                 u32_le* src_cmdbuf);
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 04cae3a43..e8ece8164 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -8,13 +8,14 @@
 #include <functional>
 #include <memory>
 #include <thread>
-#include <unordered_map>
+#include <unordered_set>
 #include <utility>
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "common/microprofile.h"
 #include "common/thread.h"
+#include "common/thread_worker.h"
 #include "core/arm/arm_interface.h"
 #include "core/arm/cpu_interrupt_handler.h"
 #include "core/arm/exclusive_monitor.h"
@@ -35,6 +36,7 @@
 #include "core/hle/kernel/physical_core.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
+#include "core/hle/kernel/service_thread.h"
 #include "core/hle/kernel/shared_memory.h"
 #include "core/hle/kernel/synchronization.h"
 #include "core/hle/kernel/thread.h"
@@ -60,6 +62,8 @@ struct KernelCore::Impl {
        RegisterHostThread();
        global_scheduler_context = std::make_unique<Kernel::GlobalSchedulerContext>(kernel);
+        service_thread_manager =
+            std::make_unique<Common::ThreadWorker>(1, "yuzu:ServiceThreadManager");
        InitializePhysicalCores();
        InitializeSystemResourceLimit(kernel);
@@ -76,6 +80,12 @@ struct KernelCore::Impl {
    }
    void Shutdown() {
+        process_list.clear();
+        // Ensures all service threads gracefully shutdown
+        service_thread_manager.reset();
+        service_threads.clear();
        next_object_id = 0;
        next_kernel_process_id = Process::InitialKIPIDMin;
        next_user_process_id = Process::ProcessIDMin;
@@ -89,8 +99,6 @@ struct KernelCore::Impl {
        cores.clear();
-        process_list.clear();
        current_process = nullptr;
        system_resource_limit = nullptr;
@@ -103,10 +111,8 @@ struct KernelCore::Impl {
        exclusive_monitor.reset();
-        num_host_threads = 0;
+        // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
-        std::fill(register_host_thread_keys.begin(), register_host_thread_keys.end(),
+        next_host_thread_id = Core::Hardware::NUM_CPU_CORES;
-                  std::thread::id{});
-        std::fill(register_host_thread_values.begin(), register_host_thread_values.end(), 0);
    }
    void InitializePhysicalCores() {
@@ -186,52 +192,46 @@ struct KernelCore::Impl {
        }
    }
+    /// Creates a new host thread ID, should only be called by GetHostThreadId
+    u32 AllocateHostThreadId(std::optional<std::size_t> core_id) {
+        if (core_id) {
+            // The first for slots are reserved for CPU core threads
+            ASSERT(*core_id < Core::Hardware::NUM_CPU_CORES);
+            return static_cast<u32>(*core_id);
+        } else {
+            return next_host_thread_id++;
+        }
+    }
+    /// Gets the host thread ID for the caller, allocating a new one if this is the first time
+    u32 GetHostThreadId(std::optional<std::size_t> core_id = std::nullopt) {
+        const thread_local auto host_thread_id{AllocateHostThreadId(core_id)};
+        return host_thread_id;
+    }
+    /// Registers a CPU core thread by allocating a host thread ID for it
    void RegisterCoreThread(std::size_t core_id) {
-        const std::thread::id this_id = std::this_thread::get_id();
+        ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+        const auto this_id = GetHostThreadId(core_id);
        if (!is_multicore) {
            single_core_thread_id = this_id;
        }
-        const auto end =
-            register_host_thread_keys.begin() + static_cast<ptrdiff_t>(num_host_threads);
-        const auto it = std::find(register_host_thread_keys.begin(), end, this_id);
-        ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
-        ASSERT(it == end);
-        InsertHostThread(static_cast<u32>(core_id));
    }
+    /// Registers a new host thread by allocating a host thread ID for it
    void RegisterHostThread() {
-        const std::thread::id this_id = std::this_thread::get_id();
+        [[maybe_unused]] const auto this_id = GetHostThreadId();
-        const auto end =
-            register_host_thread_keys.begin() + static_cast<ptrdiff_t>(num_host_threads);
-        const auto it = std::find(register_host_thread_keys.begin(), end, this_id);
-        if (it == end) {
-            InsertHostThread(registered_thread_ids++);
-        }
-    }
-    void InsertHostThread(u32 value) {
-        const size_t index = num_host_threads++;
-        ASSERT_MSG(index < NUM_REGISTRABLE_HOST_THREADS, "Too many host threads");
-        register_host_thread_values[index] = value;
-        register_host_thread_keys[index] = std::this_thread::get_id();
    }
-    [[nodiscard]] u32 GetCurrentHostThreadID() const {
+    [[nodiscard]] u32 GetCurrentHostThreadID() {
-        const std::thread::id this_id = std::this_thread::get_id();
+        const auto this_id = GetHostThreadId();
        if (!is_multicore && single_core_thread_id == this_id) {
            return static_cast<u32>(system.GetCpuManager().CurrentCore());
        }
-        const auto end =
+        return this_id;
-            register_host_thread_keys.begin() + static_cast<ptrdiff_t>(num_host_threads);
-        const auto it = std::find(register_host_thread_keys.begin(), end, this_id);
-        if (it == end) {
-            return Core::INVALID_HOST_THREAD_ID;
-        }
-        return register_host_thread_values[static_cast<size_t>(
-            std::distance(register_host_thread_keys.begin(), it))];
    }
-    Core::EmuThreadHandle GetCurrentEmuThreadID() const {
+    [[nodiscard]] Core::EmuThreadHandle GetCurrentEmuThreadID() {
        Core::EmuThreadHandle result = Core::EmuThreadHandle::InvalidHandle();
        result.host_handle = GetCurrentHostThreadID();
        if (result.host_handle >= Core::Hardware::NUM_CPU_CORES) {
@@ -325,15 +325,8 @@ struct KernelCore::Impl {
    std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
    std::vector<Kernel::PhysicalCore> cores;
-    // 0-3 IDs represent core threads, >3 represent others
+    // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others
-    std::atomic<u32> registered_thread_ids{Core::Hardware::NUM_CPU_CORES};
+    std::atomic<u32> next_host_thread_id{Core::Hardware::NUM_CPU_CORES};
-    // Number of host threads is a relatively high number to avoid overflowing
-    static constexpr size_t NUM_REGISTRABLE_HOST_THREADS = 64;
-    std::atomic<size_t> num_host_threads{0};
-    std::array<std::atomic<std::thread::id>, NUM_REGISTRABLE_HOST_THREADS>
-        register_host_thread_keys{};
-    std::array<std::atomic<u32>, NUM_REGISTRABLE_HOST_THREADS> register_host_thread_values{};
    // Kernel memory management
    std::unique_ptr<Memory::MemoryManager> memory_manager;
@@ -345,12 +338,19 @@ struct KernelCore::Impl {
    std::shared_ptr<Kernel::SharedMemory> irs_shared_mem;
    std::shared_ptr<Kernel::SharedMemory> time_shared_mem;
+    // Threads used for services
+    std::unordered_set<std::shared_ptr<Kernel::ServiceThread>> service_threads;
+    // Service threads are managed by a worker thread, so that a calling service thread can queue up
+    // the release of itself
+    std::unique_ptr<Common::ThreadWorker> service_thread_manager;
    std::array<std::shared_ptr<Thread>, Core::Hardware::NUM_CPU_CORES> suspend_threads{};
    std::array<Core::CPUInterruptHandler, Core::Hardware::NUM_CPU_CORES> interrupts{};
    std::array<std::unique_ptr<Kernel::KScheduler>, Core::Hardware::NUM_CPU_CORES> schedulers{};
    bool is_multicore{};
-    std::thread::id single_core_thread_id{};
+    u32 single_core_thread_id{};
    std::array<u64, Core::Hardware::NUM_CPU_CORES> svc_ticks{};
@@ -639,4 +639,19 @@ void KernelCore::ExitSVCProfile() {
    MicroProfileLeave(MICROPROFILE_TOKEN(Kernel_SVC), impl->svc_ticks[core]);
 }
+std::weak_ptr<Kernel::ServiceThread> KernelCore::CreateServiceThread(const std::string& name) {
+    auto service_thread = std::make_shared<Kernel::ServiceThread>(*this, 1, name);
+    impl->service_thread_manager->QueueWork(
+        [this, service_thread] { impl->service_threads.emplace(service_thread); });
+    return service_thread;
+}
+void KernelCore::ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread) {
+    impl->service_thread_manager->QueueWork([this, service_thread] {
+        if (auto strong_ptr = service_thread.lock()) {
+            impl->service_threads.erase(strong_ptr);
+        }
+    });
+}
 } // namespace Kernel
diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h
index 5846c3f39..e3169f5a7 100644
--- a/src/core/hle/kernel/kernel.h
+++ b/src/core/hle/kernel/kernel.h
@@ -42,6 +42,7 @@ class Process;
 class ResourceLimit;
 class KScheduler;
 class SharedMemory;
+class ServiceThread;
 class Synchronization;
 class Thread;
 class TimeManager;
@@ -227,6 +228,22 @@ public:
    void ExitSVCProfile();
+    /**
+     * Creates an HLE service thread, which are used to execute service routines asynchronously.
+     * While these are allocated per ServerSession, these need to be owned and managed outside of
+     * ServerSession to avoid a circular dependency.
+     * @param name String name for the ServerSession creating this thread, used for debug purposes.
+     * @returns The a weak pointer newly created service thread.
+     */
+    std::weak_ptr<Kernel::ServiceThread> CreateServiceThread(const std::string& name);
+    /**
+     * Releases a HLE service thread, instructing KernelCore to free it. This should be called when
+     * the ServerSession associated with the thread is destroyed.
+     * @param service_thread Service thread to release.
+     */
+    void ReleaseServiceThread(std::weak_ptr<Kernel::ServiceThread> service_thread);
 private:
    friend class Object;
    friend class Process;
diff --git a/src/core/hle/kernel/memory/memory_block.h b/src/core/hle/kernel/memory/memory_block.h
index 37fe19916..83acece1e 100644
--- a/src/core/hle/kernel/memory/memory_block.h
+++ b/src/core/hle/kernel/memory/memory_block.h
@@ -73,12 +73,12 @@ enum class MemoryState : u32 {
    ThreadLocal =
        static_cast<u32>(Svc::MemoryState::ThreadLocal) | FlagMapped | FlagReferenceCounted,
-    Transfered = static_cast<u32>(Svc::MemoryState::Transfered) | FlagsMisc |
+    Transferred = static_cast<u32>(Svc::MemoryState::Transferred) | FlagsMisc |
-                 FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
+                  FlagCanAlignedDeviceMap | FlagCanChangeAttribute | FlagCanUseIpc |
-                 FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+                  FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
-    SharedTransfered = static_cast<u32>(Svc::MemoryState::SharedTransfered) | FlagsMisc |
+    SharedTransferred = static_cast<u32>(Svc::MemoryState::SharedTransferred) | FlagsMisc |
-                       FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
+                        FlagCanAlignedDeviceMap | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
    SharedCode = static_cast<u32>(Svc::MemoryState::SharedCode) | FlagMapped |
                 FlagReferenceCounted | FlagCanUseNonSecureIpc | FlagCanUseNonDeviceIpc,
@@ -111,8 +111,8 @@ static_assert(static_cast<u32>(MemoryState::AliasCodeData) == 0x03FFBD09);
 static_assert(static_cast<u32>(MemoryState::Ipc) == 0x005C3C0A);
 static_assert(static_cast<u32>(MemoryState::Stack) == 0x005C3C0B);
 static_assert(static_cast<u32>(MemoryState::ThreadLocal) == 0x0040200C);
-static_assert(static_cast<u32>(MemoryState::Transfered) == 0x015C3C0D);
+static_assert(static_cast<u32>(MemoryState::Transferred) == 0x015C3C0D);
-static_assert(static_cast<u32>(MemoryState::SharedTransfered) == 0x005C380E);
+static_assert(static_cast<u32>(MemoryState::SharedTransferred) == 0x005C380E);
 static_assert(static_cast<u32>(MemoryState::SharedCode) == 0x0040380F);
 static_assert(static_cast<u32>(MemoryState::Inaccessible) == 0x00000010);
 static_assert(static_cast<u32>(MemoryState::NonSecureIpc) == 0x005C3811);
diff --git a/src/core/hle/kernel/memory/page_table.cpp b/src/core/hle/kernel/memory/page_table.cpp
index f53a7be82..080886554 100644
--- a/src/core/hle/kernel/memory/page_table.cpp
+++ b/src/core/hle/kernel/memory/page_table.cpp
@@ -265,7 +265,7 @@ ResultCode PageTable::InitializeForProcess(FileSys::ProgramAddressSpaceType as_t
    physical_memory_usage = 0;
    memory_pool = pool;
-    page_table_impl.Resize(address_space_width, PageBits, true);
+    page_table_impl.Resize(address_space_width, PageBits);
    return InitializeMemoryLayout(start, end);
 }
@@ -1007,8 +1007,8 @@ constexpr VAddr PageTable::GetRegionAddress(MemoryState state) const {
    case MemoryState::Shared:
    case MemoryState::AliasCode:
    case MemoryState::AliasCodeData:
-    case MemoryState::Transfered:
+    case MemoryState::Transferred:
-    case MemoryState::SharedTransfered:
+    case MemoryState::SharedTransferred:
    case MemoryState::SharedCode:
    case MemoryState::GeneratedCode:
    case MemoryState::CodeOut:
@@ -1042,8 +1042,8 @@ constexpr std::size_t PageTable::GetRegionSize(MemoryState state) const {
    case MemoryState::Shared:
    case MemoryState::AliasCode:
    case MemoryState::AliasCodeData:
-    case MemoryState::Transfered:
+    case MemoryState::Transferred:
-    case MemoryState::SharedTransfered:
+    case MemoryState::SharedTransferred:
    case MemoryState::SharedCode:
    case MemoryState::GeneratedCode:
    case MemoryState::CodeOut:
@@ -1080,8 +1080,8 @@ constexpr bool PageTable::CanContain(VAddr addr, std::size_t size, MemoryState s
    case MemoryState::AliasCodeData:
    case MemoryState::Stack:
    case MemoryState::ThreadLocal:
-    case MemoryState::Transfered:
+    case MemoryState::Transferred:
-    case MemoryState::SharedTransfered:
+    case MemoryState::SharedTransferred:
    case MemoryState::SharedCode:
    case MemoryState::GeneratedCode:
    case MemoryState::CodeOut:
diff --git a/src/core/hle/kernel/server_session.cpp b/src/core/hle/kernel/server_session.cpp
index a35c8aa4b..b40fe3916 100644
--- a/src/core/hle/kernel/server_session.cpp
+++ b/src/core/hle/kernel/server_session.cpp
@@ -25,19 +25,19 @@
 namespace Kernel {
 ServerSession::ServerSession(KernelCore& kernel) : SynchronizationObject{kernel} {}
-ServerSession::~ServerSession() = default;
+ServerSession::~ServerSession() {
+    kernel.ReleaseServiceThread(service_thread);
+}
 ResultVal<std::shared_ptr<ServerSession>> ServerSession::Create(KernelCore& kernel,
                                                                std::shared_ptr<Session> parent,
                                                                std::string name) {
    std::shared_ptr<ServerSession> session{std::make_shared<ServerSession>(kernel)};
-    session->request_event =
-        Core::Timing::CreateEvent(name, [session](std::uintptr_t, std::chrono::nanoseconds) {
-            session->CompleteSyncRequest();
-        });
    session->name = std::move(name);
    session->parent = std::move(parent);
+    session->service_thread = kernel.CreateServiceThread(session->name);
    return MakeResult(std::move(session));
 }
@@ -142,16 +142,16 @@ ResultCode ServerSession::QueueSyncRequest(std::shared_ptr<Thread> thread,
        std::make_shared<HLERequestContext>(kernel, memory, SharedFrom(this), std::move(thread));
    context->PopulateFromIncomingCommandBuffer(kernel.CurrentProcess()->GetHandleTable(), cmd_buf);
-    request_queue.Push(std::move(context));
+    if (auto strong_ptr = service_thread.lock()) {
+        strong_ptr->QueueSyncRequest(*this, std::move(context));
+        return RESULT_SUCCESS;
+    }
    return RESULT_SUCCESS;
 }
-ResultCode ServerSession::CompleteSyncRequest() {
+ResultCode ServerSession::CompleteSyncRequest(HLERequestContext& context) {
-    ASSERT(!request_queue.Empty());
-    auto& context = *request_queue.Front();
    ResultCode result = RESULT_SUCCESS;
    // If the session has been converted to a domain, handle the domain request
    if (IsDomain() && context.HasDomainMessageHeader()) {
@@ -177,18 +177,13 @@ ResultCode ServerSession::CompleteSyncRequest() {
        }
    }
-    request_queue.Pop();
    return result;
 }
 ResultCode ServerSession::HandleSyncRequest(std::shared_ptr<Thread> thread,
                                            Core::Memory::Memory& memory,
                                            Core::Timing::CoreTiming& core_timing) {
-    const ResultCode result = QueueSyncRequest(std::move(thread), memory);
+    return QueueSyncRequest(std::move(thread), memory);
-    const auto delay = std::chrono::nanoseconds{kernel.IsMulticore() ? 0 : 20000};
-    core_timing.ScheduleEvent(delay, request_event, {});
-    return result;
 }
 } // namespace Kernel
diff --git a/src/core/hle/kernel/server_session.h b/src/core/hle/kernel/server_session.h
index d23e9ec68..e8d1d99ea 100644
--- a/src/core/hle/kernel/server_session.h
+++ b/src/core/hle/kernel/server_session.h
@@ -10,6 +10,7 @@
 #include <vector>
 #include "common/threadsafe_queue.h"
+#include "core/hle/kernel/service_thread.h"
 #include "core/hle/kernel/synchronization_object.h"
 #include "core/hle/result.h"
@@ -43,6 +44,8 @@ class Thread;
 * TLS buffer and control is transferred back to it.
 */
 class ServerSession final : public SynchronizationObject {
+    friend class ServiceThread;
 public:
    explicit ServerSession(KernelCore& kernel);
    ~ServerSession() override;
@@ -132,7 +135,7 @@ private:
    ResultCode QueueSyncRequest(std::shared_ptr<Thread> thread, Core::Memory::Memory& memory);
    /// Completes a sync request from the emulated application.
-    ResultCode CompleteSyncRequest();
+    ResultCode CompleteSyncRequest(HLERequestContext& context);
    /// Handles a SyncRequest to a domain, forwarding the request to the proper object or closing an
    /// object handle.
@@ -163,11 +166,8 @@ private:
    /// The name of this session (optional)
    std::string name;
-    /// Core timing event used to schedule the service request at some point in the future
+    /// Thread to dispatch service requests
-    std::shared_ptr<Core::Timing::EventType> request_event;
+    std::weak_ptr<ServiceThread> service_thread;
-    /// Queue of scheduled service requests
-    Common::MPSCQueue<std::shared_ptr<Kernel::HLERequestContext>> request_queue;
 };
 } // namespace Kernel
diff --git a/src/core/hle/kernel/service_thread.cpp b/src/core/hle/kernel/service_thread.cpp
new file mode 100644
index 000000000..ee46f3e21
--- /dev/null
+++ b/src/core/hle/kernel/service_thread.cpp
@@ -0,0 +1,110 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#include <condition_variable>
+#include <functional>
+#include <mutex>
+#include <thread>
+#include <vector>
+#include <queue>
+#include "common/assert.h"
+#include "common/scope_exit.h"
+#include "common/thread.h"
+#include "core/core.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/server_session.h"
+#include "core/hle/kernel/service_thread.h"
+#include "core/hle/lock.h"
+#include "video_core/renderer_base.h"
+namespace Kernel {
+class ServiceThread::Impl final {
+public:
+    explicit Impl(KernelCore& kernel, std::size_t num_threads, const std::string& name);
+    ~Impl();
+    void QueueSyncRequest(ServerSession& session, std::shared_ptr<HLERequestContext>&& context);
+private:
+    std::vector<std::thread> threads;
+    std::queue<std::function<void()>> requests;
+    std::mutex queue_mutex;
+    std::condition_variable condition;
+    const std::string service_name;
+    bool stop{};
+};
+ServiceThread::Impl::Impl(KernelCore& kernel, std::size_t num_threads, const std::string& name)
+    : service_name{name} {
+    for (std::size_t i = 0; i < num_threads; ++i)
+        threads.emplace_back([this, &kernel] {
+            Common::SetCurrentThreadName(std::string{"yuzu:HleService:" + service_name}.c_str());
+            // Wait for first request before trying to acquire a render context
+            {
+                std::unique_lock lock{queue_mutex};
+                condition.wait(lock, [this] { return stop || !requests.empty(); });
+            }
+            kernel.RegisterHostThread();
+            while (true) {
+                std::function<void()> task;
+                {
+                    std::unique_lock lock{queue_mutex};
+                    condition.wait(lock, [this] { return stop || !requests.empty(); });
+                    if (stop || requests.empty()) {
+                        return;
+                    }
+                    task = std::move(requests.front());
+                    requests.pop();
+                }
+                task();
+            }
+        });
+}
+void ServiceThread::Impl::QueueSyncRequest(ServerSession& session,
+                                           std::shared_ptr<HLERequestContext>&& context) {
+    {
+        std::unique_lock lock{queue_mutex};
+        // ServerSession owns the service thread, so we cannot caption a strong pointer here in the
+        // event that the ServerSession is terminated.
+        std::weak_ptr<ServerSession> weak_ptr{SharedFrom(&session)};
+        requests.emplace([weak_ptr, context{std::move(context)}]() {
+            if (auto strong_ptr = weak_ptr.lock()) {
+                strong_ptr->CompleteSyncRequest(*context);
+            }
+        });
+    }
+    condition.notify_one();
+}
+ServiceThread::Impl::~Impl() {
+    {
+        std::unique_lock lock{queue_mutex};
+        stop = true;
+    }
+    condition.notify_all();
+    for (std::thread& thread : threads) {
+        thread.join();
+    }
+}
+ServiceThread::ServiceThread(KernelCore& kernel, std::size_t num_threads, const std::string& name)
+    : impl{std::make_unique<Impl>(kernel, num_threads, name)} {}
+ServiceThread::~ServiceThread() = default;
+void ServiceThread::QueueSyncRequest(ServerSession& session,
+                                     std::shared_ptr<HLERequestContext>&& context) {
+    impl->QueueSyncRequest(session, std::move(context));
+}
+} // namespace Kernel
diff --git a/src/core/hle/kernel/service_thread.h b/src/core/hle/kernel/service_thread.h
new file mode 100644
index 000000000..025ab8fb5
--- /dev/null
+++ b/src/core/hle/kernel/service_thread.h
@@ -0,0 +1,28 @@
+// Copyright 2020 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#pragma once
+#include <memory>
+#include <string>
+namespace Kernel {
+class HLERequestContext;
+class KernelCore;
+class ServerSession;
+class ServiceThread final {
+public:
+    explicit ServiceThread(KernelCore& kernel, std::size_t num_threads, const std::string& name);
+    ~ServiceThread();
+    void QueueSyncRequest(ServerSession& session, std::shared_ptr<HLERequestContext>&& context);
+private:
+    class Impl;
+    std::unique_ptr<Impl> impl;
+};
+} // namespace Kernel
diff --git a/src/core/hle/kernel/svc_types.h b/src/core/hle/kernel/svc_types.h
index 986724beb..11e1d8e2d 100644
--- a/src/core/hle/kernel/svc_types.h
+++ b/src/core/hle/kernel/svc_types.h
@@ -23,8 +23,8 @@ enum class MemoryState : u32 {
    Ipc = 0x0A,
    Stack = 0x0B,
    ThreadLocal = 0x0C,
-    Transfered = 0x0D,
+    Transferred = 0x0D,
-    SharedTransfered = 0x0E,
+    SharedTransferred = 0x0E,
    SharedCode = 0x0F,
    Inaccessible = 0x10,
    NonSecureIpc = 0x11,
diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp
index cb13210e5..c9808060a 100644
--- a/src/core/hle/service/am/am.cpp
+++ b/src/core/hle/service/am/am.cpp
@@ -560,14 +560,14 @@ void ISelfController::GetAccumulatedSuspendedTickChangedEvent(Kernel::HLERequest
 AppletMessageQueue::AppletMessageQueue(Kernel::KernelCore& kernel) {
    on_new_message =
-        Kernel::WritableEvent::CreateEventPair(kernel, "AMMessageQueue:OnMessageRecieved");
+        Kernel::WritableEvent::CreateEventPair(kernel, "AMMessageQueue:OnMessageReceived");
    on_operation_mode_changed =
        Kernel::WritableEvent::CreateEventPair(kernel, "AMMessageQueue:OperationModeChanged");
 }
 AppletMessageQueue::~AppletMessageQueue() = default;
-const std::shared_ptr<Kernel::ReadableEvent>& AppletMessageQueue::GetMesssageRecieveEvent() const {
+const std::shared_ptr<Kernel::ReadableEvent>& AppletMessageQueue::GetMessageReceiveEvent() const {
    return on_new_message.readable;
 }
@@ -675,7 +675,7 @@ void ICommonStateGetter::GetEventHandle(Kernel::HLERequestContext& ctx) {
    IPC::ResponseBuilder rb{ctx, 2, 1};
    rb.Push(RESULT_SUCCESS);
-    rb.PushCopyObjects(msg_queue->GetMesssageRecieveEvent());
+    rb.PushCopyObjects(msg_queue->GetMessageReceiveEvent());
 }
 void ICommonStateGetter::ReceiveMessage(Kernel::HLERequestContext& ctx) {
diff --git a/src/core/hle/service/am/am.h b/src/core/hle/service/am/am.h
index b1da0d081..f51aca1af 100644
--- a/src/core/hle/service/am/am.h
+++ b/src/core/hle/service/am/am.h
@@ -55,7 +55,7 @@ public:
    explicit AppletMessageQueue(Kernel::KernelCore& kernel);
    ~AppletMessageQueue();
-    const std::shared_ptr<Kernel::ReadableEvent>& GetMesssageRecieveEvent() const;
+    const std::shared_ptr<Kernel::ReadableEvent>& GetMessageReceiveEvent() const;
    const std::shared_ptr<Kernel::ReadableEvent>& GetOperationModeChangedEvent() const;
    void PushMessage(AppletMessage msg);
    AppletMessage PopMessage();
diff --git a/src/core/hle/service/audio/audout_u.cpp b/src/core/hle/service/audio/audout_u.cpp
index 145f47ee2..0cd797109 100644
--- a/src/core/hle/service/audio/audout_u.cpp
+++ b/src/core/hle/service/audio/audout_u.cpp
@@ -70,8 +70,10 @@ public:
            Kernel::WritableEvent::CreateEventPair(system.Kernel(), "IAudioOutBufferReleased");
        stream = audio_core.OpenStream(system.CoreTiming(), audio_params.sample_rate,
-                                       audio_params.channel_count, std::move(unique_name),
+                                       audio_params.channel_count, std::move(unique_name), [this] {
-                                       [this] { buffer_event.writable->Signal(); });
+                                           const auto guard = LockService();
+                                           buffer_event.writable->Signal();
+                                       });
    }
 private:
diff --git a/src/core/hle/service/audio/audren_u.cpp b/src/core/hle/service/audio/audren_u.cpp
index 6e7b7316c..c5c22d053 100644
--- a/src/core/hle/service/audio/audren_u.cpp
+++ b/src/core/hle/service/audio/audren_u.cpp
@@ -49,16 +49,16 @@ public:
        system_event =
            Kernel::WritableEvent::CreateEventPair(system.Kernel(), "IAudioRenderer:SystemEvent");
-        renderer = std::make_unique<AudioCore::AudioRenderer>(system.CoreTiming(), system.Memory(),
+        renderer = std::make_unique<AudioCore::AudioRenderer>(
-                                                              audren_params, system_event.writable,
+            system.CoreTiming(), system.Memory(), audren_params,
-                                                              instance_number);
+            [this]() {
+                const auto guard = LockService();
+                system_event.writable->Signal();
+            },
+            instance_number);
    }
 private:
-    void UpdateAudioCallback() {
-        system_event.writable->Signal();
-    }
    void GetSampleRate(Kernel::HLERequestContext& ctx) {
        LOG_DEBUG(Service_Audio, "called");
diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp
index b3c7234e1..8d95f74e6 100644
--- a/src/core/hle/service/hid/hid.cpp
+++ b/src/core/hle/service/hid/hid.cpp
@@ -78,11 +78,13 @@ IAppletResource::IAppletResource(Core::System& system_)
    pad_update_event = Core::Timing::CreateEvent(
        "HID::UpdatePadCallback",
        [this](std::uintptr_t user_data, std::chrono::nanoseconds ns_late) {
+            const auto guard = LockService();
            UpdateControllers(user_data, ns_late);
        });
    motion_update_event = Core::Timing::CreateEvent(
        "HID::MotionPadCallback",
        [this](std::uintptr_t user_data, std::chrono::nanoseconds ns_late) {
+            const auto guard = LockService();
            UpdateMotion(user_data, ns_late);
        });
diff --git a/src/core/hle/service/nvdrv/devices/nvdevice.h b/src/core/hle/service/nvdrv/devices/nvdevice.h
index 44a8bc060..5681599ba 100644
--- a/src/core/hle/service/nvdrv/devices/nvdevice.h
+++ b/src/core/hle/service/nvdrv/devices/nvdevice.h
@@ -31,8 +31,8 @@ public:
     * @param output A buffer where the output data will be written to.
     * @returns The result code of the ioctl.
     */
-    virtual NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    virtual NvResult Ioctl1(Ioctl command, const std::vector<u8>& input,
-                            IoctlCtrl& ctrl) = 0;
+                            std::vector<u8>& output) = 0;
    /**
     * Handles an ioctl2 request.
@@ -43,8 +43,7 @@ public:
     * @returns The result code of the ioctl.
     */
    virtual NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                            const std::vector<u8>& inline_input, std::vector<u8>& output,
+                            const std::vector<u8>& inline_input, std::vector<u8>& output) = 0;
-                            IoctlCtrl& ctrl) = 0;
    /**
     * Handles an ioctl3 request.
@@ -55,7 +54,7 @@ public:
     * @returns The result code of the ioctl.
     */
    virtual NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                            std::vector<u8>& inline_output, IoctlCtrl& ctrl) = 0;
+                            std::vector<u8>& inline_output) = 0;
 protected:
    Core::System& system;
diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
index 170a7c9a0..ce615c758 100644
--- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp
@@ -18,21 +18,20 @@ nvdisp_disp0::nvdisp_disp0(Core::System& system, std::shared_ptr<nvmap> nvmap_de
    : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
 nvdisp_disp0 ::~nvdisp_disp0() = default;
-NvResult nvdisp_disp0::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvdisp_disp0::Ioctl1(Ioctl command, const std::vector<u8>& input,
-                              IoctlCtrl& ctrl) {
+                              std::vector<u8>& output) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvdisp_disp0::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                              const std::vector<u8>& inline_input, std::vector<u8>& output,
+                              const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                              IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvdisp_disp0::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                              std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                              std::vector<u8>& inline_output) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
diff --git a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
index eb7575e40..55a33b7e4 100644
--- a/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
+++ b/src/core/hle/service/nvdrv/devices/nvdisp_disp0.h
@@ -20,13 +20,11 @@ public:
    explicit nvdisp_disp0(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
    ~nvdisp_disp0() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
    /// Performs a screen flip, drawing the buffer pointed to by the handle.
    void flip(u32 buffer_handle, u32 offset, u32 format, u32 width, u32 height, u32 stride,
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
index 4e0652c39..6b062e10e 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.cpp
@@ -21,8 +21,8 @@ nvhost_as_gpu::nvhost_as_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_
    : nvdevice(system), nvmap_dev(std::move(nvmap_dev)) {}
 nvhost_as_gpu::~nvhost_as_gpu() = default;
-NvResult nvhost_as_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvhost_as_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input,
-                               IoctlCtrl& ctrl) {
+                               std::vector<u8>& output) {
    switch (command.group) {
    case 'A':
        switch (command.cmd) {
@@ -55,14 +55,13 @@ NvResult nvhost_as_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input, std:
 }
 NvResult nvhost_as_gpu::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                               const std::vector<u8>& inline_input, std::vector<u8>& output,
+                               const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                               IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvhost_as_gpu::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                               std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                               std::vector<u8>& inline_output) {
    switch (command.group) {
    case 'A':
        switch (command.cmd) {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h
index 2bd355af9..08035fa0e 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h
@@ -30,13 +30,11 @@ public:
    explicit nvhost_as_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
    ~nvhost_as_gpu() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
 private:
    class BufferMap final {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
index 92d31b620..fea3b7b9f 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp
@@ -20,8 +20,7 @@ nvhost_ctrl::nvhost_ctrl(Core::System& system, EventInterface& events_interface,
    : nvdevice(system), events_interface{events_interface}, syncpoint_manager{syncpoint_manager} {}
 nvhost_ctrl::~nvhost_ctrl() = default;
-NvResult nvhost_ctrl::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvhost_ctrl::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
-                             IoctlCtrl& ctrl) {
    switch (command.group) {
    case 0x0:
        switch (command.cmd) {
@@ -30,9 +29,9 @@ NvResult nvhost_ctrl::Ioctl1(Ioctl command, const std::vector<u8>& input, std::v
        case 0x1c:
            return IocCtrlClearEventWait(input, output);
        case 0x1d:
-            return IocCtrlEventWait(input, output, false, ctrl);
+            return IocCtrlEventWait(input, output, false);
        case 0x1e:
-            return IocCtrlEventWait(input, output, true, ctrl);
+            return IocCtrlEventWait(input, output, true);
        case 0x1f:
            return IocCtrlEventRegister(input, output);
        case 0x20:
@@ -48,14 +47,13 @@ NvResult nvhost_ctrl::Ioctl1(Ioctl command, const std::vector<u8>& input, std::v
 }
 NvResult nvhost_ctrl::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                             const std::vector<u8>& inline_input, std::vector<u8>& output,
+                             const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                             IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvhost_ctrl::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                             std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                             std::vector<u8>& inline_outpu) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
@@ -69,7 +67,7 @@ NvResult nvhost_ctrl::NvOsGetConfigU32(const std::vector<u8>& input, std::vector
 }
 NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output,
-                                       bool is_async, IoctlCtrl& ctrl) {
+                                       bool is_async) {
    IocCtrlEventWaitParams params{};
    std::memcpy(&params, input.data(), sizeof(params));
    LOG_DEBUG(Service_NVDRV, "syncpt_id={}, threshold={}, timeout={}, is_async={}",
@@ -141,12 +139,6 @@ NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector
        params.value |= event_id;
        event.event.writable->Clear();
        gpu.RegisterSyncptInterrupt(params.syncpt_id, target_value);
-        if (!is_async && ctrl.fresh_call) {
-            ctrl.must_delay = true;
-            ctrl.timeout = params.timeout;
-            ctrl.event_id = event_id;
-            return NvResult::Timeout;
-        }
        std::memcpy(output.data(), &params, sizeof(params));
        return NvResult::Timeout;
    }
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
index 107168e21..c5aa1362a 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl.h
@@ -18,13 +18,11 @@ public:
                         SyncpointManager& syncpoint_manager);
    ~nvhost_ctrl() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
 private:
    struct IocSyncptReadParams {
@@ -123,8 +121,7 @@ private:
    static_assert(sizeof(IocCtrlEventKill) == 8, "IocCtrlEventKill is incorrect size");
    NvResult NvOsGetConfigU32(const std::vector<u8>& input, std::vector<u8>& output);
-    NvResult IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output, bool is_async,
+    NvResult IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output, bool is_async);
-                              IoctlCtrl& ctrl);
    NvResult IocCtrlEventRegister(const std::vector<u8>& input, std::vector<u8>& output);
    NvResult IocCtrlEventUnregister(const std::vector<u8>& input, std::vector<u8>& output);
    NvResult IocCtrlClearEventWait(const std::vector<u8>& input, std::vector<u8>& output);
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
index 647f5907e..0320d3ae2 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.cpp
@@ -16,7 +16,7 @@ nvhost_ctrl_gpu::nvhost_ctrl_gpu(Core::System& system) : nvdevice(system) {}
 nvhost_ctrl_gpu::~nvhost_ctrl_gpu() = default;
 NvResult nvhost_ctrl_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input,
-                                 std::vector<u8>& output, IoctlCtrl& ctrl) {
+                                 std::vector<u8>& output) {
    switch (command.group) {
    case 'G':
        switch (command.cmd) {
@@ -48,15 +48,13 @@ NvResult nvhost_ctrl_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input,
 }
 NvResult nvhost_ctrl_gpu::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                                 const std::vector<u8>& inline_input, std::vector<u8>& output,
+                                 const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                                 IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvhost_ctrl_gpu::Ioctl3(Ioctl command, const std::vector<u8>& input,
-                                 std::vector<u8>& output, std::vector<u8>& inline_output,
+                                 std::vector<u8>& output, std::vector<u8>& inline_output) {
-                                 IoctlCtrl& ctrl) {
    switch (command.group) {
    case 'G':
        switch (command.cmd) {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h
index c2fffe734..137b88238 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_ctrl_gpu.h
@@ -16,13 +16,11 @@ public:
    explicit nvhost_ctrl_gpu(Core::System& system);
    ~nvhost_ctrl_gpu() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
 private:
    struct IoctlGpuCharacteristics {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
index b0c2caba5..af8b3d9f1 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
@@ -23,8 +23,7 @@ nvhost_gpu::nvhost_gpu(Core::System& system, std::shared_ptr<nvmap> nvmap_dev,
 nvhost_gpu::~nvhost_gpu() = default;
-NvResult nvhost_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvhost_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
-                            IoctlCtrl& ctrl) {
    switch (command.group) {
    case 0x0:
        switch (command.cmd) {
@@ -76,8 +75,7 @@ NvResult nvhost_gpu::Ioctl1(Ioctl command, const std::vector<u8>& input, std::ve
 };
 NvResult nvhost_gpu::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                            const std::vector<u8>& inline_input, std::vector<u8>& output,
+                            const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                            IoctlCtrl& ctrl) {
    switch (command.group) {
    case 'H':
        switch (command.cmd) {
@@ -91,7 +89,7 @@ NvResult nvhost_gpu::Ioctl2(Ioctl command, const std::vector<u8>& input,
 }
 NvResult nvhost_gpu::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                            std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                            std::vector<u8>& inline_output) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
index aa0048a9d..e0298b4fe 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
@@ -26,13 +26,11 @@ public:
                        SyncpointManager& syncpoint_manager);
    ~nvhost_gpu() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
 private:
    enum class CtxObjects : u32_le {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
index b8328c314..d8735491c 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp
@@ -15,8 +15,8 @@ nvhost_nvdec::nvhost_nvdec(Core::System& system, std::shared_ptr<nvmap> nvmap_de
    : nvhost_nvdec_common(system, std::move(nvmap_dev)) {}
 nvhost_nvdec::~nvhost_nvdec() = default;
-NvResult nvhost_nvdec::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvhost_nvdec::Ioctl1(Ioctl command, const std::vector<u8>& input,
-                              IoctlCtrl& ctrl) {
+                              std::vector<u8>& output) {
    switch (command.group) {
    case 0x0:
        switch (command.cmd) {
@@ -58,14 +58,13 @@ NvResult nvhost_nvdec::Ioctl1(Ioctl command, const std::vector<u8>& input, std::
 }
 NvResult nvhost_nvdec::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                              const std::vector<u8>& inline_input, std::vector<u8>& output,
+                              const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                              IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvhost_nvdec::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                              std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                              std::vector<u8>& inline_output) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
index 884ed6c5b..79b8b6de1 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.h
@@ -14,13 +14,11 @@ public:
    explicit nvhost_nvdec(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
    ~nvhost_nvdec() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
 };
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp
index 6f4ab0ab3..2d06955c0 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.cpp
@@ -13,8 +13,8 @@ namespace Service::Nvidia::Devices {
 nvhost_nvjpg::nvhost_nvjpg(Core::System& system) : nvdevice(system) {}
 nvhost_nvjpg::~nvhost_nvjpg() = default;
-NvResult nvhost_nvjpg::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvhost_nvjpg::Ioctl1(Ioctl command, const std::vector<u8>& input,
-                              IoctlCtrl& ctrl) {
+                              std::vector<u8>& output) {
    switch (command.group) {
    case 'H':
        switch (command.cmd) {
@@ -33,14 +33,13 @@ NvResult nvhost_nvjpg::Ioctl1(Ioctl command, const std::vector<u8>& input, std::
 }
 NvResult nvhost_nvjpg::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                              const std::vector<u8>& inline_input, std::vector<u8>& output,
+                              const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                              IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvhost_nvjpg::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                              std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                              std::vector<u8>& inline_output) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h
index 6fb99d959..43948d18d 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvjpg.h
@@ -16,13 +16,11 @@ public:
    explicit nvhost_nvjpg(Core::System& system);
    ~nvhost_nvjpg() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
 private:
    struct IoctlSetNvmapFD {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
index 55a17f423..805fe86ae 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp
@@ -15,8 +15,7 @@ nvhost_vic::nvhost_vic(Core::System& system, std::shared_ptr<nvmap> nvmap_dev)
 nvhost_vic::~nvhost_vic() = default;
-NvResult nvhost_vic::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvhost_vic::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
-                            IoctlCtrl& ctrl) {
    switch (command.group) {
    case 0x0:
        switch (command.cmd) {
@@ -51,14 +50,13 @@ NvResult nvhost_vic::Ioctl1(Ioctl command, const std::vector<u8>& input, std::ve
 }
 NvResult nvhost_vic::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                            const std::vector<u8>& inline_input, std::vector<u8>& output,
+                            const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                            IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvhost_vic::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                            std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                            std::vector<u8>& inline_output) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.h b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
index 7f4858cd4..b2e11f4d4 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_vic.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.h
@@ -14,12 +14,10 @@ public:
    explicit nvhost_vic(Core::System& system, std::shared_ptr<nvmap> nvmap_dev);
    ~nvhost_vic();
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
 };
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvmap.cpp b/src/core/hle/service/nvdrv/devices/nvmap.cpp
index 910cfee51..4015a2740 100644
--- a/src/core/hle/service/nvdrv/devices/nvmap.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvmap.cpp
@@ -19,8 +19,7 @@ nvmap::nvmap(Core::System& system) : nvdevice(system) {
 nvmap::~nvmap() = default;
-NvResult nvmap::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+NvResult nvmap::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) {
-                       IoctlCtrl& ctrl) {
    switch (command.group) {
    case 0x1:
        switch (command.cmd) {
@@ -49,14 +48,13 @@ NvResult nvmap::Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<
 }
 NvResult nvmap::Ioctl2(Ioctl command, const std::vector<u8>& input,
-                       const std::vector<u8>& inline_input, std::vector<u8>& output,
+                       const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                       IoctlCtrl& ctrl) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
 NvResult nvmap::Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                       std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                       std::vector<u8>& inline_output) {
    UNIMPLEMENTED_MSG("Unimplemented ioctl={:08X}", command.raw);
    return NvResult::NotImplemented;
 }
diff --git a/src/core/hle/service/nvdrv/devices/nvmap.h b/src/core/hle/service/nvdrv/devices/nvmap.h
index c0c2fa5eb..4484bd79f 100644
--- a/src/core/hle/service/nvdrv/devices/nvmap.h
+++ b/src/core/hle/service/nvdrv/devices/nvmap.h
@@ -19,13 +19,11 @@ public:
    explicit nvmap(Core::System& system);
    ~nvmap() override;
-    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
+    NvResult Ioctl1(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl2(Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output,
+                    const std::vector<u8>& inline_input, std::vector<u8>& output) override;
-                    IoctlCtrl& ctrl) override;
    NvResult Ioctl3(Ioctl command, const std::vector<u8>& input, std::vector<u8>& output,
-                    std::vector<u8>& inline_output, IoctlCtrl& ctrl) override;
+                    std::vector<u8>& inline_output) override;
    /// Returns the allocated address of an nvmap object given its handle.
    VAddr GetObjectAddress(u32 handle) const;
diff --git a/src/core/hle/service/nvdrv/interface.cpp b/src/core/hle/service/nvdrv/interface.cpp
index d72c531f6..cc23b001c 100644
--- a/src/core/hle/service/nvdrv/interface.cpp
+++ b/src/core/hle/service/nvdrv/interface.cpp
@@ -61,32 +61,9 @@ void NVDRV::Ioctl1(Kernel::HLERequestContext& ctx) {
    std::vector<u8> output_buffer(ctx.GetWriteBufferSize(0));
    const auto input_buffer = ctx.ReadBuffer(0);
-    IoctlCtrl ctrl{};
+    const auto nv_result = nvdrv->Ioctl1(fd, command, input_buffer, output_buffer);
+    if (command.is_out != 0) {
-    const auto nv_result = nvdrv->Ioctl1(fd, command, input_buffer, output_buffer, ctrl);
+        ctx.WriteBuffer(output_buffer);
-    if (ctrl.must_delay) {
-        ctrl.fresh_call = false;
-        ctx.SleepClientThread(
-            "NVServices::DelayedResponse", ctrl.timeout,
-            [=, this](std::shared_ptr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx_,
-                      Kernel::ThreadWakeupReason reason) {
-                IoctlCtrl ctrl2{ctrl};
-                std::vector<u8> tmp_output = output_buffer;
-                const auto nv_result2 = nvdrv->Ioctl1(fd, command, input_buffer, tmp_output, ctrl2);
-                if (command.is_out != 0) {
-                    ctx.WriteBuffer(tmp_output);
-                }
-                IPC::ResponseBuilder rb{ctx_, 3};
-                rb.Push(RESULT_SUCCESS);
-                rb.PushEnum(nv_result2);
-            },
-            nvdrv->GetEventWriteable(ctrl.event_id));
-    } else {
-        if (command.is_out != 0) {
-            ctx.WriteBuffer(output_buffer);
-        }
    }
    IPC::ResponseBuilder rb{ctx, 3};
@@ -110,36 +87,8 @@ void NVDRV::Ioctl2(Kernel::HLERequestContext& ctx) {
    const auto input_inlined_buffer = ctx.ReadBuffer(1);
    std::vector<u8> output_buffer(ctx.GetWriteBufferSize(0));
-    IoctlCtrl ctrl{};
    const auto nv_result =
-        nvdrv->Ioctl2(fd, command, input_buffer, input_inlined_buffer, output_buffer, ctrl);
+        nvdrv->Ioctl2(fd, command, input_buffer, input_inlined_buffer, output_buffer);
-    if (ctrl.must_delay) {
-        ctrl.fresh_call = false;
-        ctx.SleepClientThread(
-            "NVServices::DelayedResponse", ctrl.timeout,
-            [=, this](std::shared_ptr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx_,
-                      Kernel::ThreadWakeupReason reason) {
-                IoctlCtrl ctrl2{ctrl};
-                std::vector<u8> tmp_output = output_buffer;
-                const auto nv_result2 = nvdrv->Ioctl2(fd, command, input_buffer,
-                                                      input_inlined_buffer, tmp_output, ctrl2);
-                if (command.is_out != 0) {
-                    ctx.WriteBuffer(tmp_output);
-                }
-                IPC::ResponseBuilder rb{ctx_, 3};
-                rb.Push(RESULT_SUCCESS);
-                rb.PushEnum(nv_result2);
-            },
-            nvdrv->GetEventWriteable(ctrl.event_id));
-    } else {
-        if (command.is_out != 0) {
-            ctx.WriteBuffer(output_buffer);
-        }
-    }
    if (command.is_out != 0) {
        ctx.WriteBuffer(output_buffer);
    }
@@ -165,36 +114,11 @@ void NVDRV::Ioctl3(Kernel::HLERequestContext& ctx) {
    std::vector<u8> output_buffer(ctx.GetWriteBufferSize(0));
    std::vector<u8> output_buffer_inline(ctx.GetWriteBufferSize(1));
-    IoctlCtrl ctrl{};
    const auto nv_result =
-        nvdrv->Ioctl3(fd, command, input_buffer, output_buffer, output_buffer_inline, ctrl);
+        nvdrv->Ioctl3(fd, command, input_buffer, output_buffer, output_buffer_inline);
-    if (ctrl.must_delay) {
+    if (command.is_out != 0) {
-        ctrl.fresh_call = false;
+        ctx.WriteBuffer(output_buffer, 0);
-        ctx.SleepClientThread(
+        ctx.WriteBuffer(output_buffer_inline, 1);
-            "NVServices::DelayedResponse", ctrl.timeout,
-            [=, this](std::shared_ptr<Kernel::Thread> thread, Kernel::HLERequestContext& ctx_,
-                      Kernel::ThreadWakeupReason reason) {
-                IoctlCtrl ctrl2{ctrl};
-                std::vector<u8> tmp_output = output_buffer;
-                std::vector<u8> tmp_output2 = output_buffer;
-                const auto nv_result2 =
-                    nvdrv->Ioctl3(fd, command, input_buffer, tmp_output, tmp_output2, ctrl2);
-                if (command.is_out != 0) {
-                    ctx.WriteBuffer(tmp_output, 0);
-                    ctx.WriteBuffer(tmp_output2, 1);
-                }
-                IPC::ResponseBuilder rb{ctx_, 3};
-                rb.Push(RESULT_SUCCESS);
-                rb.PushEnum(nv_result2);
-            },
-            nvdrv->GetEventWriteable(ctrl.event_id));
-    } else {
-        if (command.is_out != 0) {
-            ctx.WriteBuffer(output_buffer, 0);
-            ctx.WriteBuffer(output_buffer_inline, 1);
-        }
    }
    IPC::ResponseBuilder rb{ctx, 3};
diff --git a/src/core/hle/service/nvdrv/nvdata.h b/src/core/hle/service/nvdrv/nvdata.h
index a3c4ecd85..3294bc0e7 100644
--- a/src/core/hle/service/nvdrv/nvdata.h
+++ b/src/core/hle/service/nvdrv/nvdata.h
@@ -97,15 +97,4 @@ union Ioctl {
    BitField<31, 1, u32> is_out;
 };
-struct IoctlCtrl {
-    // First call done to the servioce for services that call itself again after a call.
-    bool fresh_call{true};
-    // Tells the Ioctl Wrapper that it must delay the IPC response and send the thread to sleep
-    bool must_delay{};
-    // Timeout for the delay
-    s64 timeout{};
-    // NV Event Id
-    s32 event_id{-1};
-};
 } // namespace Service::Nvidia
diff --git a/src/core/hle/service/nvdrv/nvdrv.cpp b/src/core/hle/service/nvdrv/nvdrv.cpp
index 8e0c9f093..e03195afe 100644
--- a/src/core/hle/service/nvdrv/nvdrv.cpp
+++ b/src/core/hle/service/nvdrv/nvdrv.cpp
@@ -91,7 +91,7 @@ DeviceFD Module::Open(const std::string& device_name) {
 }
 NvResult Module::Ioctl1(DeviceFD fd, Ioctl command, const std::vector<u8>& input,
-                        std::vector<u8>& output, IoctlCtrl& ctrl) {
+                        std::vector<u8>& output) {
    if (fd < 0) {
        LOG_ERROR(Service_NVDRV, "Invalid DeviceFD={}!", fd);
        return NvResult::InvalidState;
@@ -104,12 +104,11 @@ NvResult Module::Ioctl1(DeviceFD fd, Ioctl command, const std::vector<u8>& input
        return NvResult::NotImplemented;
    }
-    return itr->second->Ioctl1(command, input, output, ctrl);
+    return itr->second->Ioctl1(command, input, output);
 }
 NvResult Module::Ioctl2(DeviceFD fd, Ioctl command, const std::vector<u8>& input,
-                        const std::vector<u8>& inline_input, std::vector<u8>& output,
+                        const std::vector<u8>& inline_input, std::vector<u8>& output) {
-                        IoctlCtrl& ctrl) {
    if (fd < 0) {
        LOG_ERROR(Service_NVDRV, "Invalid DeviceFD={}!", fd);
        return NvResult::InvalidState;
@@ -122,11 +121,11 @@ NvResult Module::Ioctl2(DeviceFD fd, Ioctl command, const std::vector<u8>& input
        return NvResult::NotImplemented;
    }
-    return itr->second->Ioctl2(command, input, inline_input, output, ctrl);
+    return itr->second->Ioctl2(command, input, inline_input, output);
 }
 NvResult Module::Ioctl3(DeviceFD fd, Ioctl command, const std::vector<u8>& input,
-                        std::vector<u8>& output, std::vector<u8>& inline_output, IoctlCtrl& ctrl) {
+                        std::vector<u8>& output, std::vector<u8>& inline_output) {
    if (fd < 0) {
        LOG_ERROR(Service_NVDRV, "Invalid DeviceFD={}!", fd);
        return NvResult::InvalidState;
@@ -139,7 +138,7 @@ NvResult Module::Ioctl3(DeviceFD fd, Ioctl command, const std::vector<u8>& input
        return NvResult::NotImplemented;
    }
-    return itr->second->Ioctl3(command, input, output, inline_output, ctrl);
+    return itr->second->Ioctl3(command, input, output, inline_output);
 }
 NvResult Module::Close(DeviceFD fd) {
diff --git a/src/core/hle/service/nvdrv/nvdrv.h b/src/core/hle/service/nvdrv/nvdrv.h
index 5985d2179..144e657e5 100644
--- a/src/core/hle/service/nvdrv/nvdrv.h
+++ b/src/core/hle/service/nvdrv/nvdrv.h
@@ -119,13 +119,13 @@ public:
    /// Sends an ioctl command to the specified file descriptor.
    NvResult Ioctl1(DeviceFD fd, Ioctl command, const std::vector<u8>& input,
-                    std::vector<u8>& output, IoctlCtrl& ctrl);
+                    std::vector<u8>& output);
    NvResult Ioctl2(DeviceFD fd, Ioctl command, const std::vector<u8>& input,
-                    const std::vector<u8>& inline_input, std::vector<u8>& output, IoctlCtrl& ctrl);
+                    const std::vector<u8>& inline_input, std::vector<u8>& output);
    NvResult Ioctl3(DeviceFD fd, Ioctl command, const std::vector<u8>& input,
-                    std::vector<u8>& output, std::vector<u8>& inline_output, IoctlCtrl& ctrl);
+                    std::vector<u8>& output, std::vector<u8>& inline_output);
    /// Closes a device file descriptor and returns operation success.
    NvResult Close(DeviceFD fd);
diff --git a/src/core/hle/service/nvflinger/buffer_queue.cpp b/src/core/hle/service/nvflinger/buffer_queue.cpp
index 377f47e8e..0e6bde9f5 100644
--- a/src/core/hle/service/nvflinger/buffer_queue.cpp
+++ b/src/core/hle/service/nvflinger/buffer_queue.cpp
@@ -25,7 +25,12 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer)
    ASSERT(slot < buffer_slots);
    LOG_WARNING(Service, "Adding graphics buffer {}", slot);
-    free_buffers.push_back(slot);
+    {
+        std::unique_lock lock{queue_mutex};
+        free_buffers.push_back(slot);
+    }
+    condition.notify_one();
    buffers[slot] = {
        .slot = slot,
        .status = Buffer::Status::Free,
@@ -41,10 +46,20 @@ void BufferQueue::SetPreallocatedBuffer(u32 slot, const IGBPBuffer& igbp_buffer)
 std::optional<std::pair<u32, Service::Nvidia::MultiFence*>> BufferQueue::DequeueBuffer(u32 width,
                                                                                       u32 height) {
+    // Wait for first request before trying to dequeue
+    {
+        std::unique_lock lock{queue_mutex};
+        condition.wait(lock, [this] { return !free_buffers.empty() || !is_connect; });
+    }
-    if (free_buffers.empty()) {
+    if (!is_connect) {
+        // Buffer was disconnected while the thread was blocked, this is most likely due to
+        // emulation being stopped
        return std::nullopt;
    }
+    std::unique_lock lock{queue_mutex};
    auto f_itr = free_buffers.begin();
    auto slot = buffers.size();
@@ -97,7 +112,11 @@ void BufferQueue::CancelBuffer(u32 slot, const Service::Nvidia::MultiFence& mult
    buffers[slot].multi_fence = multi_fence;
    buffers[slot].swap_interval = 0;
-    free_buffers.push_back(slot);
+    {
+        std::unique_lock lock{queue_mutex};
+        free_buffers.push_back(slot);
+    }
+    condition.notify_one();
    buffer_wait_event.writable->Signal();
 }
@@ -127,15 +146,26 @@ void BufferQueue::ReleaseBuffer(u32 slot) {
    ASSERT(buffers[slot].slot == slot);
    buffers[slot].status = Buffer::Status::Free;
-    free_buffers.push_back(slot);
+    {
+        std::unique_lock lock{queue_mutex};
+        free_buffers.push_back(slot);
+    }
+    condition.notify_one();
    buffer_wait_event.writable->Signal();
 }
+void BufferQueue::Connect() {
+    queue_sequence.clear();
+    is_connect = true;
+}
 void BufferQueue::Disconnect() {
    buffers.fill({});
    queue_sequence.clear();
    buffer_wait_event.writable->Signal();
+    is_connect = false;
+    condition.notify_one();
 }
 u32 BufferQueue::Query(QueryType type) {
diff --git a/src/core/hle/service/nvflinger/buffer_queue.h b/src/core/hle/service/nvflinger/buffer_queue.h
index e610923cb..a2f60d9eb 100644
--- a/src/core/hle/service/nvflinger/buffer_queue.h
+++ b/src/core/hle/service/nvflinger/buffer_queue.h
@@ -4,7 +4,9 @@
 #pragma once
+#include <condition_variable>
 #include <list>
+#include <mutex>
 #include <optional>
 #include <vector>
@@ -99,6 +101,7 @@ public:
    void CancelBuffer(u32 slot, const Service::Nvidia::MultiFence& multi_fence);
    std::optional<std::reference_wrapper<const Buffer>> AcquireBuffer();
    void ReleaseBuffer(u32 slot);
+    void Connect();
    void Disconnect();
    u32 Query(QueryType type);
@@ -106,18 +109,28 @@ public:
        return id;
    }
+    bool IsConnected() const {
+        return is_connect;
+    }
    std::shared_ptr<Kernel::WritableEvent> GetWritableBufferWaitEvent() const;
    std::shared_ptr<Kernel::ReadableEvent> GetBufferWaitEvent() const;
 private:
-    u32 id;
+    BufferQueue(const BufferQueue&) = delete;
-    u64 layer_id;
+    u32 id{};
+    u64 layer_id{};
+    std::atomic_bool is_connect{};
    std::list<u32> free_buffers;
    std::array<Buffer, buffer_slots> buffers;
    std::list<u32> queue_sequence;
    Kernel::EventPair buffer_wait_event;
+    std::mutex queue_mutex;
+    std::condition_variable condition;
 };
 } // namespace Service::NVFlinger
diff --git a/src/core/hle/service/nvflinger/nvflinger.cpp b/src/core/hle/service/nvflinger/nvflinger.cpp
index 44aa2bdae..4b3581949 100644
--- a/src/core/hle/service/nvflinger/nvflinger.cpp
+++ b/src/core/hle/service/nvflinger/nvflinger.cpp
@@ -88,6 +88,10 @@ NVFlinger::NVFlinger(Core::System& system) : system(system) {
 }
 NVFlinger::~NVFlinger() {
+    for (auto& buffer_queue : buffer_queues) {
+        buffer_queue->Disconnect();
+    }
    if (system.IsMulticore()) {
        is_running = false;
        wait_event->Set();
@@ -104,6 +108,8 @@ void NVFlinger::SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance) {
 }
 std::optional<u64> NVFlinger::OpenDisplay(std::string_view name) {
+    const auto guard = Lock();
    LOG_DEBUG(Service, "Opening \"{}\" display", name);
    // TODO(Subv): Currently we only support the Default display.
@@ -121,6 +127,7 @@ std::optional<u64> NVFlinger::OpenDisplay(std::string_view name) {
 }
 std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
+    const auto guard = Lock();
    auto* const display = FindDisplay(display_id);
    if (display == nullptr) {
@@ -129,18 +136,22 @@ std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
    const u64 layer_id = next_layer_id++;
    const u32 buffer_queue_id = next_buffer_queue_id++;
-    buffer_queues.emplace_back(system.Kernel(), buffer_queue_id, layer_id);
+    buffer_queues.emplace_back(
-    display->CreateLayer(layer_id, buffer_queues.back());
+        std::make_unique<BufferQueue>(system.Kernel(), buffer_queue_id, layer_id));
+    display->CreateLayer(layer_id, *buffer_queues.back());
    return layer_id;
 }
 void NVFlinger::CloseLayer(u64 layer_id) {
+    const auto guard = Lock();
    for (auto& display : displays) {
        display.CloseLayer(layer_id);
    }
 }
 std::optional<u32> NVFlinger::FindBufferQueueId(u64 display_id, u64 layer_id) const {
+    const auto guard = Lock();
    const auto* const layer = FindLayer(display_id, layer_id);
    if (layer == nullptr) {
@@ -151,6 +162,7 @@ std::optional<u32> NVFlinger::FindBufferQueueId(u64 display_id, u64 layer_id) co
 }
 std::shared_ptr<Kernel::ReadableEvent> NVFlinger::FindVsyncEvent(u64 display_id) const {
+    const auto guard = Lock();
    auto* const display = FindDisplay(display_id);
    if (display == nullptr) {
@@ -160,20 +172,16 @@ std::shared_ptr<Kernel::ReadableEvent> NVFlinger::FindVsyncEvent(u64 display_id)
    return display->GetVSyncEvent();
 }
-BufferQueue& NVFlinger::FindBufferQueue(u32 id) {
+BufferQueue* NVFlinger::FindBufferQueue(u32 id) {
+    const auto guard = Lock();
    const auto itr = std::find_if(buffer_queues.begin(), buffer_queues.end(),
-                                  [id](const auto& queue) { return queue.GetId() == id; });
+                                  [id](const auto& queue) { return queue->GetId() == id; });
-    ASSERT(itr != buffer_queues.end());
+    if (itr == buffer_queues.end()) {
-    return *itr;
+        return nullptr;
-}
+    }
-const BufferQueue& NVFlinger::FindBufferQueue(u32 id) const {
-    const auto itr = std::find_if(buffer_queues.begin(), buffer_queues.end(),
-                                  [id](const auto& queue) { return queue.GetId() == id; });
-    ASSERT(itr != buffer_queues.end());
+    return itr->get();
-    return *itr;
 }
 VI::Display* NVFlinger::FindDisplay(u64 display_id) {
diff --git a/src/core/hle/service/nvflinger/nvflinger.h b/src/core/hle/service/nvflinger/nvflinger.h
index 1ebe949c0..c6765259f 100644
--- a/src/core/hle/service/nvflinger/nvflinger.h
+++ b/src/core/hle/service/nvflinger/nvflinger.h
@@ -75,10 +75,7 @@ public:
    [[nodiscard]] std::shared_ptr<Kernel::ReadableEvent> FindVsyncEvent(u64 display_id) const;
    /// Obtains a buffer queue identified by the ID.
-    [[nodiscard]] BufferQueue& FindBufferQueue(u32 id);
+    [[nodiscard]] BufferQueue* FindBufferQueue(u32 id);
-    /// Obtains a buffer queue identified by the ID.
-    [[nodiscard]] const BufferQueue& FindBufferQueue(u32 id) const;
    /// Performs a composition request to the emulated nvidia GPU and triggers the vsync events when
    /// finished.
@@ -86,11 +83,11 @@ public:
    [[nodiscard]] s64 GetNextTicks() const;
+private:
    [[nodiscard]] std::unique_lock<std::mutex> Lock() const {
        return std::unique_lock{*guard};
    }
-private:
    /// Finds the display identified by the specified ID.
    [[nodiscard]] VI::Display* FindDisplay(u64 display_id);
@@ -110,7 +107,7 @@ private:
    std::shared_ptr<Nvidia::Module> nvdrv;
    std::vector<VI::Display> displays;
-    std::vector<BufferQueue> buffer_queues;
+    std::vector<std::unique_ptr<BufferQueue>> buffer_queues;
    /// Id to use for the next layer that is created, this counter is shared among all displays.
    u64 next_layer_id = 1;
diff --git a/src/core/hle/service/service.cpp b/src/core/hle/service/service.cpp
index abf3d1ea3..ff2a5b1db 100644
--- a/src/core/hle/service/service.cpp
+++ b/src/core/hle/service/service.cpp
@@ -95,9 +95,14 @@ ServiceFrameworkBase::ServiceFrameworkBase(Core::System& system_, const char* se
    : system{system_}, service_name{service_name_}, max_sessions{max_sessions_},
      handler_invoker{handler_invoker_} {}
-ServiceFrameworkBase::~ServiceFrameworkBase() = default;
+ServiceFrameworkBase::~ServiceFrameworkBase() {
+    // Wait for other threads to release access before destroying
+    const auto guard = LockService();
+}
 void ServiceFrameworkBase::InstallAsService(SM::ServiceManager& service_manager) {
+    const auto guard = LockService();
    ASSERT(!port_installed);
    auto port = service_manager.RegisterService(service_name, max_sessions).Unwrap();
@@ -106,6 +111,8 @@ void ServiceFrameworkBase::InstallAsService(SM::ServiceManager& service_manager)
 }
 void ServiceFrameworkBase::InstallAsNamedPort(Kernel::KernelCore& kernel) {
+    const auto guard = LockService();
    ASSERT(!port_installed);
    auto [server_port, client_port] =
@@ -115,17 +122,6 @@ void ServiceFrameworkBase::InstallAsNamedPort(Kernel::KernelCore& kernel) {
    port_installed = true;
 }
-std::shared_ptr<Kernel::ClientPort> ServiceFrameworkBase::CreatePort(Kernel::KernelCore& kernel) {
-    ASSERT(!port_installed);
-    auto [server_port, client_port] =
-        Kernel::ServerPort::CreatePortPair(kernel, max_sessions, service_name);
-    auto port = MakeResult(std::move(server_port)).Unwrap();
-    port->SetHleHandler(shared_from_this());
-    port_installed = true;
-    return client_port;
-}
 void ServiceFrameworkBase::RegisterHandlersBase(const FunctionInfoBase* functions, std::size_t n) {
    handlers.reserve(handlers.size() + n);
    for (std::size_t i = 0; i < n; ++i) {
@@ -164,6 +160,8 @@ void ServiceFrameworkBase::InvokeRequest(Kernel::HLERequestContext& ctx) {
 }
 ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& context) {
+    const auto guard = LockService();
    switch (context.GetCommandType()) {
    case IPC::CommandType::Close: {
        IPC::ResponseBuilder rb{context, 2};
@@ -184,7 +182,11 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
        UNIMPLEMENTED_MSG("command_type={}", context.GetCommandType());
    }
-    context.WriteToOutgoingCommandBuffer(context.GetThread());
+    // If emulation was shutdown, we are closing service threads, do not write the response back to
+    // memory that may be shutting down as well.
+    if (system.IsPoweredOn()) {
+        context.WriteToOutgoingCommandBuffer(context.GetThread());
+    }
    return RESULT_SUCCESS;
 }
diff --git a/src/core/hle/service/service.h b/src/core/hle/service/service.h
index 62a182310..916445517 100644
--- a/src/core/hle/service/service.h
+++ b/src/core/hle/service/service.h
@@ -5,9 +5,11 @@
 #pragma once
 #include <cstddef>
+#include <mutex>
 #include <string>
 #include <boost/container/flat_map.hpp>
 #include "common/common_types.h"
+#include "common/spin_lock.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/kernel/object.h"
@@ -68,11 +70,9 @@ public:
    void InstallAsService(SM::ServiceManager& service_manager);
    /// Creates a port pair and registers it on the kernel's global port registry.
    void InstallAsNamedPort(Kernel::KernelCore& kernel);
-    /// Creates and returns an unregistered port for the service.
+    /// Invokes a service request routine.
-    std::shared_ptr<Kernel::ClientPort> CreatePort(Kernel::KernelCore& kernel);
    void InvokeRequest(Kernel::HLERequestContext& ctx);
+    /// Handles a synchronization request for the service.
    ResultCode HandleSyncRequest(Kernel::HLERequestContext& context) override;
 protected:
@@ -80,6 +80,11 @@ protected:
    template <typename Self>
    using HandlerFnP = void (Self::*)(Kernel::HLERequestContext&);
+    /// Used to gain exclusive access to the service members, e.g. from CoreTiming thread.
+    [[nodiscard]] std::scoped_lock<Common::SpinLock> LockService() {
+        return std::scoped_lock{lock_service};
+    }
    /// System context that the service operates under.
    Core::System& system;
@@ -115,6 +120,9 @@ private:
    /// Function used to safely up-cast pointers to the derived class before invoking a handler.
    InvokerFn* handler_invoker;
    boost::container::flat_map<u32, FunctionInfoBase> handlers;
+    /// Used to gain exclusive access to the service members, e.g. from CoreTiming thread.
+    Common::SpinLock lock_service;
 };
 /**
diff --git a/src/core/hle/service/sockets/blocking_worker.h b/src/core/hle/service/sockets/blocking_worker.h
deleted file mode 100644
index 2d53e52b6..000000000
--- a/src/core/hle/service/sockets/blocking_worker.h
+++ /dev/null
@@ -1,161 +0,0 @@
-// Copyright 2020 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#pragma once
-#include <atomic>
-#include <memory>
-#include <string>
-#include <string_view>
-#include <thread>
-#include <variant>
-#include <vector>
-#include <fmt/format.h>
-#include "common/assert.h"
-#include "common/microprofile.h"
-#include "common/thread.h"
-#include "core/core.h"
-#include "core/hle/kernel/hle_ipc.h"
-#include "core/hle/kernel/kernel.h"
-#include "core/hle/kernel/thread.h"
-#include "core/hle/kernel/writable_event.h"
-namespace Service::Sockets {
-/**
- * Worker abstraction to execute blocking calls on host without blocking the guest thread
- *
- * @tparam Service  Service where the work is executed
- * @tparam Types Types of work to execute
- */
-template <class Service, class... Types>
-class BlockingWorker {
-    using This = BlockingWorker<Service, Types...>;
-    using WorkVariant = std::variant<std::monostate, Types...>;
-public:
-    /// Create a new worker
-    static std::unique_ptr<This> Create(Core::System& system, Service* service,
-                                        std::string_view name) {
-        return std::unique_ptr<This>(new This(system, service, name));
-    }
-    ~BlockingWorker() {
-        while (!is_available.load(std::memory_order_relaxed)) {
-            // Busy wait until work is finished
-            std::this_thread::yield();
-        }
-        // Monostate means to exit the thread
-        work = std::monostate{};
-        work_event.Set();
-        thread.join();
-    }
-    /**
-     * Try to capture the worker to send work after a success
-     * @returns True when the worker has been successfully captured
-     */
-    bool TryCapture() {
-        bool expected = true;
-        return is_available.compare_exchange_weak(expected, false, std::memory_order_relaxed,
-                                                  std::memory_order_relaxed);
-    }
-    /**
-     * Send work to this worker abstraction
-     * @see TryCapture must be called before attempting to call this function
-     */
-    template <class Work>
-    void SendWork(Work new_work) {
-        ASSERT_MSG(!is_available, "Trying to send work on a worker that's not captured");
-        work = std::move(new_work);
-        work_event.Set();
-    }
-    /// Generate a callback for @see SleepClientThread
-    template <class Work>
-    auto Callback() {
-        return [this](std::shared_ptr<Kernel::Thread>, Kernel::HLERequestContext& ctx,
-                      Kernel::ThreadWakeupReason reason) {
-            ASSERT(reason == Kernel::ThreadWakeupReason::Signal);
-            std::get<Work>(work).Response(ctx);
-            is_available.store(true);
-        };
-    }
-    /// Get kernel event that will be signalled by the worker when the host operation finishes
-    std::shared_ptr<Kernel::WritableEvent> KernelEvent() const {
-        return kernel_event;
-    }
-private:
-    explicit BlockingWorker(Core::System& system, Service* service, std::string_view name) {
-        auto pair = Kernel::WritableEvent::CreateEventPair(system.Kernel(), std::string(name));
-        kernel_event = std::move(pair.writable);
-        thread = std::thread([this, &system, service, name] { Run(system, service, name); });
-    }
-    void Run(Core::System& system, Service* service, std::string_view name) {
-        system.RegisterHostThread();
-        const std::string thread_name = fmt::format("yuzu:{}", name);
-        MicroProfileOnThreadCreate(thread_name.c_str());
-        Common::SetCurrentThreadName(thread_name.c_str());
-        bool keep_running = true;
-        while (keep_running) {
-            work_event.Wait();
-            const auto visit_fn = [service, &keep_running]<typename T>(T&& w) {
-                if constexpr (std::is_same_v<std::decay_t<T>, std::monostate>) {
-                    keep_running = false;
-                } else {
-                    w.Execute(service);
-                }
-            };
-            std::visit(visit_fn, work);
-            kernel_event->Signal();
-        }
-    }
-    std::thread thread;
-    WorkVariant work;
-    Common::Event work_event;
-    std::shared_ptr<Kernel::WritableEvent> kernel_event;
-    std::atomic_bool is_available{true};
-};
-template <class Service, class... Types>
-class BlockingWorkerPool {
-    using Worker = BlockingWorker<Service, Types...>;
-public:
-    explicit BlockingWorkerPool(Core::System& system_, Service* service_)
-        : system{system_}, service{service_} {}
-    /// Returns a captured worker thread, creating new ones if necessary
-    Worker* CaptureWorker() {
-        for (auto& worker : workers) {
-            if (worker->TryCapture()) {
-                return worker.get();
-            }
-        }
-        auto new_worker = Worker::Create(system, service, fmt::format("BSD:{}", workers.size()));
-        [[maybe_unused]] const bool success = new_worker->TryCapture();
-        ASSERT(success);
-        return workers.emplace_back(std::move(new_worker)).get();
-    }
-private:
-    Core::System& system;
-    Service* const service;
-    std::vector<std::unique_ptr<Worker>> workers;
-};
-} // namespace Service::Sockets
diff --git a/src/core/hle/service/sockets/bsd.cpp b/src/core/hle/service/sockets/bsd.cpp
index 67b419503..2b824059d 100644
--- a/src/core/hle/service/sockets/bsd.cpp
+++ b/src/core/hle/service/sockets/bsd.cpp
@@ -178,13 +178,12 @@ void BSD::Poll(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. nfds={} timeout={}", nfds, timeout);
-    ExecuteWork(ctx, "BSD:Poll", timeout != 0,
+    ExecuteWork(ctx, PollWork{
-                PollWork{
+                         .nfds = nfds,
-                    .nfds = nfds,
+                         .timeout = timeout,
-                    .timeout = timeout,
+                         .read_buffer = ctx.ReadBuffer(),
-                    .read_buffer = ctx.ReadBuffer(),
+                         .write_buffer = std::vector<u8>(ctx.GetWriteBufferSize()),
-                    .write_buffer = std::vector<u8>(ctx.GetWriteBufferSize()),
+                     });
-                });
 }
 void BSD::Accept(Kernel::HLERequestContext& ctx) {
@@ -193,11 +192,10 @@ void BSD::Accept(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. fd={}", fd);
-    ExecuteWork(ctx, "BSD:Accept", IsBlockingSocket(fd),
+    ExecuteWork(ctx, AcceptWork{
-                AcceptWork{
+                         .fd = fd,
-                    .fd = fd,
+                         .write_buffer = std::vector<u8>(ctx.GetWriteBufferSize()),
-                    .write_buffer = std::vector<u8>(ctx.GetWriteBufferSize()),
+                     });
-                });
 }
 void BSD::Bind(Kernel::HLERequestContext& ctx) {
@@ -215,11 +213,10 @@ void BSD::Connect(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. fd={} addrlen={}", fd, ctx.GetReadBufferSize());
-    ExecuteWork(ctx, "BSD:Connect", IsBlockingSocket(fd),
+    ExecuteWork(ctx, ConnectWork{
-                ConnectWork{
+                         .fd = fd,
-                    .fd = fd,
+                         .addr = ctx.ReadBuffer(),
-                    .addr = ctx.ReadBuffer(),
+                     });
-                });
 }
 void BSD::GetPeerName(Kernel::HLERequestContext& ctx) {
@@ -327,12 +324,11 @@ void BSD::Recv(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. fd={} flags=0x{:x} len={}", fd, flags, ctx.GetWriteBufferSize());
-    ExecuteWork(ctx, "BSD:Recv", IsBlockingSocket(fd),
+    ExecuteWork(ctx, RecvWork{
-                RecvWork{
+                         .fd = fd,
-                    .fd = fd,
+                         .flags = flags,
-                    .flags = flags,
+                         .message = std::vector<u8>(ctx.GetWriteBufferSize()),
-                    .message = std::vector<u8>(ctx.GetWriteBufferSize()),
+                     });
-                });
 }
 void BSD::RecvFrom(Kernel::HLERequestContext& ctx) {
@@ -344,13 +340,12 @@ void BSD::RecvFrom(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. fd={} flags=0x{:x} len={} addrlen={}", fd, flags,
              ctx.GetWriteBufferSize(0), ctx.GetWriteBufferSize(1));
-    ExecuteWork(ctx, "BSD:RecvFrom", IsBlockingSocket(fd),
+    ExecuteWork(ctx, RecvFromWork{
-                RecvFromWork{
+                         .fd = fd,
-                    .fd = fd,
+                         .flags = flags,
-                    .flags = flags,
+                         .message = std::vector<u8>(ctx.GetWriteBufferSize(0)),
-                    .message = std::vector<u8>(ctx.GetWriteBufferSize(0)),
+                         .addr = std::vector<u8>(ctx.GetWriteBufferSize(1)),
-                    .addr = std::vector<u8>(ctx.GetWriteBufferSize(1)),
+                     });
-                });
 }
 void BSD::Send(Kernel::HLERequestContext& ctx) {
@@ -361,12 +356,11 @@ void BSD::Send(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. fd={} flags=0x{:x} len={}", fd, flags, ctx.GetReadBufferSize());
-    ExecuteWork(ctx, "BSD:Send", IsBlockingSocket(fd),
+    ExecuteWork(ctx, SendWork{
-                SendWork{
+                         .fd = fd,
-                    .fd = fd,
+                         .flags = flags,
-                    .flags = flags,
+                         .message = ctx.ReadBuffer(),
-                    .message = ctx.ReadBuffer(),
+                     });
-                });
 }
 void BSD::SendTo(Kernel::HLERequestContext& ctx) {
@@ -377,13 +371,12 @@ void BSD::SendTo(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. fd={} flags=0x{} len={} addrlen={}", fd, flags,
              ctx.GetReadBufferSize(0), ctx.GetReadBufferSize(1));
-    ExecuteWork(ctx, "BSD:SendTo", IsBlockingSocket(fd),
+    ExecuteWork(ctx, SendToWork{
-                SendToWork{
+                         .fd = fd,
-                    .fd = fd,
+                         .flags = flags,
-                    .flags = flags,
+                         .message = ctx.ReadBuffer(0),
-                    .message = ctx.ReadBuffer(0),
+                         .addr = ctx.ReadBuffer(1),
-                    .addr = ctx.ReadBuffer(1),
+                     });
-                });
 }
 void BSD::Write(Kernel::HLERequestContext& ctx) {
@@ -392,12 +385,11 @@ void BSD::Write(Kernel::HLERequestContext& ctx) {
    LOG_DEBUG(Service, "called. fd={} len={}", fd, ctx.GetReadBufferSize());
-    ExecuteWork(ctx, "BSD:Write", IsBlockingSocket(fd),
+    ExecuteWork(ctx, SendWork{
-                SendWork{
+                         .fd = fd,
-                    .fd = fd,
+                         .flags = 0,
-                    .flags = 0,
+                         .message = ctx.ReadBuffer(),
-                    .message = ctx.ReadBuffer(),
+                     });
-                });
 }
 void BSD::Close(Kernel::HLERequestContext& ctx) {
@@ -410,24 +402,9 @@ void BSD::Close(Kernel::HLERequestContext& ctx) {
 }
 template <typename Work>
-void BSD::ExecuteWork(Kernel::HLERequestContext& ctx, std::string_view sleep_reason,
+void BSD::ExecuteWork(Kernel::HLERequestContext& ctx, Work work) {
-                      bool is_blocking, Work work) {
+    work.Execute(this);
-    if (!is_blocking) {
-        work.Execute(this);
-        work.Response(ctx);
-        return;
-    }
-    // Signal a dummy response to make IPC validation happy
-    // This will be overwritten by the SleepClientThread callback
    work.Response(ctx);
-    auto worker = worker_pool.CaptureWorker();
-    ctx.SleepClientThread(std::string(sleep_reason), std::numeric_limits<u64>::max(),
-                          worker->Callback<Work>(), worker->KernelEvent());
-    worker->SendWork(std::move(work));
 }
 std::pair<s32, Errno> BSD::SocketImpl(Domain domain, Type type, Protocol protocol) {
@@ -807,18 +784,6 @@ bool BSD::IsFileDescriptorValid(s32 fd) const noexcept {
    return true;
 }
-bool BSD::IsBlockingSocket(s32 fd) const noexcept {
-    // Inform invalid sockets as non-blocking
-    // This way we avoid using a worker thread as it will fail without blocking host
-    if (fd > static_cast<s32>(MAX_FD) || fd < 0) {
-        return false;
-    }
-    if (!file_descriptors[fd]) {
-        return false;
-    }
-    return (file_descriptors[fd]->flags & FLAG_O_NONBLOCK) != 0;
-}
 void BSD::BuildErrnoResponse(Kernel::HLERequestContext& ctx, Errno bsd_errno) const noexcept {
    IPC::ResponseBuilder rb{ctx, 4};
@@ -827,8 +792,7 @@ void BSD::BuildErrnoResponse(Kernel::HLERequestContext& ctx, Errno bsd_errno) co
    rb.PushEnum(bsd_errno);
 }
-BSD::BSD(Core::System& system_, const char* name)
+BSD::BSD(Core::System& system_, const char* name) : ServiceFramework{system_, name} {
-    : ServiceFramework{system_, name}, worker_pool{system_, this} {
    // clang-format off
    static const FunctionInfo functions[] = {
        {0, &BSD::RegisterClient, "RegisterClient"},
diff --git a/src/core/hle/service/sockets/bsd.h b/src/core/hle/service/sockets/bsd.h
index f14713fc4..6da0bfeb2 100644
--- a/src/core/hle/service/sockets/bsd.h
+++ b/src/core/hle/service/sockets/bsd.h
@@ -11,7 +11,6 @@
 #include "common/common_types.h"
 #include "core/hle/kernel/hle_ipc.h"
 #include "core/hle/service/service.h"
-#include "core/hle/service/sockets/blocking_worker.h"
 #include "core/hle/service/sockets/sockets.h"
 namespace Core {
@@ -138,8 +137,7 @@ private:
    void Close(Kernel::HLERequestContext& ctx);
    template <typename Work>
-    void ExecuteWork(Kernel::HLERequestContext& ctx, std::string_view sleep_reason,
+    void ExecuteWork(Kernel::HLERequestContext& ctx, Work work);
-                     bool is_blocking, Work work);
    std::pair<s32, Errno> SocketImpl(Domain domain, Type type, Protocol protocol);
    std::pair<s32, Errno> PollImpl(std::vector<u8>& write_buffer, std::vector<u8> read_buffer,
@@ -163,15 +161,10 @@ private:
    s32 FindFreeFileDescriptorHandle() noexcept;
    bool IsFileDescriptorValid(s32 fd) const noexcept;
-    bool IsBlockingSocket(s32 fd) const noexcept;
    void BuildErrnoResponse(Kernel::HLERequestContext& ctx, Errno bsd_errno) const noexcept;
    std::array<std::optional<FileDescriptor>, MAX_FD> file_descriptors;
-    BlockingWorkerPool<BSD, PollWork, AcceptWork, ConnectWork, RecvWork, RecvFromWork, SendWork,
-                       SendToWork>
-        worker_pool;
 };
 class BSDCFG final : public ServiceFramework<BSDCFG> {
diff --git a/src/core/hle/service/vi/vi.cpp b/src/core/hle/service/vi/vi.cpp
index 45cfffe06..968cd16b6 100644
--- a/src/core/hle/service/vi/vi.cpp
+++ b/src/core/hle/service/vi/vi.cpp
@@ -536,8 +536,7 @@ private:
        LOG_DEBUG(Service_VI, "called. id=0x{:08X} transaction={:X}, flags=0x{:08X}", id,
                  transaction, flags);
-        const auto guard = nv_flinger.Lock();
+        auto& buffer_queue = *nv_flinger.FindBufferQueue(id);
-        auto& buffer_queue = nv_flinger.FindBufferQueue(id);
        switch (transaction) {
        case TransactionId::Connect: {
@@ -547,6 +546,9 @@ private:
                                 Settings::values.resolution_factor.GetValue()),
                static_cast<u32>(static_cast<u32>(DisplayResolution::UndockedHeight) *
                                 Settings::values.resolution_factor.GetValue())};
+            buffer_queue.Connect();
            ctx.WriteBuffer(response.Serialize());
            break;
        }
@@ -563,40 +565,25 @@ private:
            IGBPDequeueBufferRequestParcel request{ctx.ReadBuffer()};
            const u32 width{request.data.width};
            const u32 height{request.data.height};
-            auto result = buffer_queue.DequeueBuffer(width, height);
+            do {
-            if (result) {
+                if (auto result = buffer_queue.DequeueBuffer(width, height); result) {
-                // Buffer is available
+                    // Buffer is available
-                IGBPDequeueBufferResponseParcel response{result->first, *result->second};
+                    IGBPDequeueBufferResponseParcel response{result->first, *result->second};
-                ctx.WriteBuffer(response.Serialize());
+                    ctx.WriteBuffer(response.Serialize());
-            } else {
+                    break;
-                // Wait the current thread until a buffer becomes available
+                }
-                ctx.SleepClientThread(
+            } while (buffer_queue.IsConnected());
-                    "IHOSBinderDriver::DequeueBuffer", UINT64_MAX,
-                    [=, this](std::shared_ptr<Kernel::Thread> thread,
-                              Kernel::HLERequestContext& ctx, Kernel::ThreadWakeupReason reason) {
-                        // Repeat TransactParcel DequeueBuffer when a buffer is available
-                        const auto guard = nv_flinger.Lock();
-                        auto& buffer_queue = nv_flinger.FindBufferQueue(id);
-                        auto result = buffer_queue.DequeueBuffer(width, height);
-                        ASSERT_MSG(result != std::nullopt, "Could not dequeue buffer.");
-                        IGBPDequeueBufferResponseParcel response{result->first, *result->second};
-                        ctx.WriteBuffer(response.Serialize());
-                        IPC::ResponseBuilder rb{ctx, 2};
-                        rb.Push(RESULT_SUCCESS);
-                    },
-                    buffer_queue.GetWritableBufferWaitEvent());
-            }
            break;
        }
        case TransactionId::RequestBuffer: {
            IGBPRequestBufferRequestParcel request{ctx.ReadBuffer()};
            auto& buffer = buffer_queue.RequestBuffer(request.slot);
            IGBPRequestBufferResponseParcel response{buffer};
            ctx.WriteBuffer(response.Serialize());
            break;
        }
        case TransactionId::QueueBuffer: {
@@ -682,7 +669,7 @@ private:
        LOG_WARNING(Service_VI, "(STUBBED) called id={}, unknown={:08X}", id, unknown);
-        const auto& buffer_queue = nv_flinger.FindBufferQueue(id);
+        const auto& buffer_queue = *nv_flinger.FindBufferQueue(id);
        // TODO(Subv): Find out what this actually is.
        IPC::ResponseBuilder rb{ctx, 2, 1};
diff --git a/src/core/memory.cpp b/src/core/memory.cpp
index 54a848936..11609682a 100644
--- a/src/core/memory.cpp
+++ b/src/core/memory.cpp
@@ -4,7 +4,6 @@
 #include <algorithm>
 #include <cstring>
-#include <mutex>
 #include <optional>
 #include <utility>
@@ -45,44 +44,16 @@ struct Memory::Impl {
        MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, target, Common::PageType::Memory);
    }
-    void MapIoRegion(Common::PageTable& page_table, VAddr base, u64 size,
-                     Common::MemoryHookPointer mmio_handler) {
-        UNIMPLEMENTED();
-    }
    void UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size) {
        ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: {:016X}", size);
        ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: {:016X}", base);
        MapPages(page_table, base / PAGE_SIZE, size / PAGE_SIZE, 0, Common::PageType::Unmapped);
    }
-    void AddDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
-                      Common::MemoryHookPointer hook) {
-        UNIMPLEMENTED();
-    }
-    void RemoveDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
-                         Common::MemoryHookPointer hook) {
-        UNIMPLEMENTED();
-    }
    bool IsValidVirtualAddress(const Kernel::Process& process, const VAddr vaddr) const {
        const auto& page_table = process.PageTable().PageTableImpl();
+        const auto [pointer, type] = page_table.pointers[vaddr >> PAGE_BITS].PointerType();
-        const u8* const page_pointer = page_table.pointers[vaddr >> PAGE_BITS];
+        return pointer != nullptr || type == Common::PageType::RasterizerCachedMemory;
-        if (page_pointer != nullptr) {
-            return true;
-        }
-        if (page_table.attributes[vaddr >> PAGE_BITS] == Common::PageType::RasterizerCachedMemory) {
-            return true;
-        }
-        if (page_table.attributes[vaddr >> PAGE_BITS] != Common::PageType::Special) {
-            return false;
-        }
-        return false;
    }
    bool IsValidVirtualAddress(VAddr vaddr) const {
@@ -100,17 +71,15 @@ struct Memory::Impl {
    }
    u8* GetPointer(const VAddr vaddr) const {
-        u8* const page_pointer{current_page_table->pointers[vaddr >> PAGE_BITS]};
+        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (page_pointer) {
+        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
-            return page_pointer + vaddr;
+            return pointer + vaddr;
        }
+        const auto type = Common::PageTable::PageInfo::ExtractType(raw_pointer);
-        if (current_page_table->attributes[vaddr >> PAGE_BITS] ==
+        if (type == Common::PageType::RasterizerCachedMemory) {
-            Common::PageType::RasterizerCachedMemory) {
            return GetPointerFromRasterizerCachedMemory(vaddr);
        }
+        return nullptr;
-        return {};
    }
    u8 Read8(const VAddr addr) {
@@ -222,7 +191,8 @@ struct Memory::Impl {
                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-            switch (page_table.attributes[page_index]) {
+            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
+            switch (type) {
            case Common::PageType::Unmapped: {
                LOG_ERROR(HW_Memory,
                          "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
@@ -231,10 +201,8 @@ struct Memory::Impl {
                break;
            }
            case Common::PageType::Memory: {
-                DEBUG_ASSERT(page_table.pointers[page_index]);
+                DEBUG_ASSERT(pointer);
+                const u8* const src_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                const u8* const src_ptr =
-                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                std::memcpy(dest_buffer, src_ptr, copy_amount);
                break;
            }
@@ -268,7 +236,8 @@ struct Memory::Impl {
                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-            switch (page_table.attributes[page_index]) {
+            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
+            switch (type) {
            case Common::PageType::Unmapped: {
                LOG_ERROR(HW_Memory,
                          "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
@@ -277,10 +246,8 @@ struct Memory::Impl {
                break;
            }
            case Common::PageType::Memory: {
-                DEBUG_ASSERT(page_table.pointers[page_index]);
+                DEBUG_ASSERT(pointer);
+                const u8* const src_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                const u8* const src_ptr =
-                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                std::memcpy(dest_buffer, src_ptr, copy_amount);
                break;
            }
@@ -320,7 +287,8 @@ struct Memory::Impl {
                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-            switch (page_table.attributes[page_index]) {
+            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
+            switch (type) {
            case Common::PageType::Unmapped: {
                LOG_ERROR(HW_Memory,
                          "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
@@ -328,10 +296,8 @@ struct Memory::Impl {
                break;
            }
            case Common::PageType::Memory: {
-                DEBUG_ASSERT(page_table.pointers[page_index]);
+                DEBUG_ASSERT(pointer);
+                u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                u8* const dest_ptr =
-                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                std::memcpy(dest_ptr, src_buffer, copy_amount);
                break;
            }
@@ -364,7 +330,8 @@ struct Memory::Impl {
                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-            switch (page_table.attributes[page_index]) {
+            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
+            switch (type) {
            case Common::PageType::Unmapped: {
                LOG_ERROR(HW_Memory,
                          "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
@@ -372,10 +339,8 @@ struct Memory::Impl {
                break;
            }
            case Common::PageType::Memory: {
-                DEBUG_ASSERT(page_table.pointers[page_index]);
+                DEBUG_ASSERT(pointer);
+                u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                u8* const dest_ptr =
-                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                std::memcpy(dest_ptr, src_buffer, copy_amount);
                break;
            }
@@ -414,7 +379,8 @@ struct Memory::Impl {
                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-            switch (page_table.attributes[page_index]) {
+            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
+            switch (type) {
            case Common::PageType::Unmapped: {
                LOG_ERROR(HW_Memory,
                          "Unmapped ZeroBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
@@ -422,10 +388,8 @@ struct Memory::Impl {
                break;
            }
            case Common::PageType::Memory: {
-                DEBUG_ASSERT(page_table.pointers[page_index]);
+                DEBUG_ASSERT(pointer);
+                u8* const dest_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                u8* dest_ptr =
-                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                std::memset(dest_ptr, 0, copy_amount);
                break;
            }
@@ -461,7 +425,8 @@ struct Memory::Impl {
                std::min(static_cast<std::size_t>(PAGE_SIZE) - page_offset, remaining_size);
            const auto current_vaddr = static_cast<VAddr>((page_index << PAGE_BITS) + page_offset);
-            switch (page_table.attributes[page_index]) {
+            const auto [pointer, type] = page_table.pointers[page_index].PointerType();
+            switch (type) {
            case Common::PageType::Unmapped: {
                LOG_ERROR(HW_Memory,
                          "Unmapped CopyBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
@@ -470,9 +435,8 @@ struct Memory::Impl {
                break;
            }
            case Common::PageType::Memory: {
-                DEBUG_ASSERT(page_table.pointers[page_index]);
+                DEBUG_ASSERT(pointer);
-                const u8* src_ptr =
+                const u8* src_ptr = pointer + page_offset + (page_index << PAGE_BITS);
-                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                WriteBlock(process, dest_addr, src_ptr, copy_amount);
                break;
            }
@@ -498,34 +462,19 @@ struct Memory::Impl {
        return CopyBlock(*system.CurrentProcess(), dest_addr, src_addr, size);
    }
-    struct PageEntry {
-        u8* const pointer;
-        const Common::PageType attribute;
-    };
-    PageEntry SafePageEntry(std::size_t base) const {
-        std::lock_guard lock{rasterizer_cache_guard};
-        return {
-            .pointer = current_page_table->pointers[base],
-            .attribute = current_page_table->attributes[base],
-        };
-    }
    void RasterizerMarkRegionCached(VAddr vaddr, u64 size, bool cached) {
-        std::lock_guard lock{rasterizer_cache_guard};
        if (vaddr == 0) {
            return;
        }
        // Iterate over a contiguous CPU address space, which corresponds to the specified GPU
        // address space, marking the region as un/cached. The region is marked un/cached at a
        // granularity of CPU pages, hence why we iterate on a CPU page basis (note: GPU page size
        // is different). This assumes the specified GPU address region is contiguous as well.
-        u64 num_pages = ((vaddr + size - 1) >> PAGE_BITS) - (vaddr >> PAGE_BITS) + 1;
+        const u64 num_pages = ((vaddr + size - 1) >> PAGE_BITS) - (vaddr >> PAGE_BITS) + 1;
-        for (unsigned i = 0; i < num_pages; ++i, vaddr += PAGE_SIZE) {
+        for (u64 i = 0; i < num_pages; ++i, vaddr += PAGE_SIZE) {
-            Common::PageType& page_type{current_page_table->attributes[vaddr >> PAGE_BITS]};
+            const Common::PageType page_type{
+                current_page_table->pointers[vaddr >> PAGE_BITS].Type()};
            if (cached) {
                // Switch page type to cached if now cached
                switch (page_type) {
@@ -534,8 +483,8 @@ struct Memory::Impl {
                    // space, for example, a system module need not have a VRAM mapping.
                    break;
                case Common::PageType::Memory:
-                    page_type = Common::PageType::RasterizerCachedMemory;
+                    current_page_table->pointers[vaddr >> PAGE_BITS].Store(
-                    current_page_table->pointers[vaddr >> PAGE_BITS] = nullptr;
+                        nullptr, Common::PageType::RasterizerCachedMemory);
                    break;
                case Common::PageType::RasterizerCachedMemory:
                    // There can be more than one GPU region mapped per CPU region, so it's common
@@ -556,16 +505,16 @@ struct Memory::Impl {
                    // that this area is already unmarked as cached.
                    break;
                case Common::PageType::RasterizerCachedMemory: {
-                    u8* pointer{GetPointerFromRasterizerCachedMemory(vaddr & ~PAGE_MASK)};
+                    u8* const pointer{GetPointerFromRasterizerCachedMemory(vaddr & ~PAGE_MASK)};
                    if (pointer == nullptr) {
                        // It's possible that this function has been called while updating the
                        // pagetable after unmapping a VMA. In that case the underlying VMA will no
                        // longer exist, and we should just leave the pagetable entry blank.
-                        page_type = Common::PageType::Unmapped;
+                        current_page_table->pointers[vaddr >> PAGE_BITS].Store(
+                            nullptr, Common::PageType::Unmapped);
                    } else {
-                        current_page_table->pointers[vaddr >> PAGE_BITS] =
+                        current_page_table->pointers[vaddr >> PAGE_BITS].Store(
-                            pointer - (vaddr & ~PAGE_MASK);
+                            pointer - (vaddr & ~PAGE_MASK), Common::PageType::Memory);
-                        page_type = Common::PageType::Memory;
                    }
                    break;
                }
@@ -595,7 +544,7 @@ struct Memory::Impl {
            auto& gpu = system.GPU();
            for (u64 i = 0; i < size; i++) {
                const auto page = base + i;
-                if (page_table.attributes[page] == Common::PageType::RasterizerCachedMemory) {
+                if (page_table.pointers[page].Type() == Common::PageType::RasterizerCachedMemory) {
                    gpu.FlushAndInvalidateRegion(page << PAGE_BITS, PAGE_SIZE);
                }
            }
@@ -610,20 +559,18 @@ struct Memory::Impl {
                       "Mapping memory page without a pointer @ {:016x}", base * PAGE_SIZE);
            while (base != end) {
-                page_table.attributes[base] = type;
+                page_table.pointers[base].Store(nullptr, type);
-                page_table.pointers[base] = nullptr;
                page_table.backing_addr[base] = 0;
                base += 1;
            }
        } else {
            while (base != end) {
-                page_table.pointers[base] =
+                page_table.pointers[base].Store(
-                    system.DeviceMemory().GetPointer(target) - (base << PAGE_BITS);
+                    system.DeviceMemory().GetPointer(target) - (base << PAGE_BITS), type);
-                page_table.attributes[base] = type;
                page_table.backing_addr[base] = target - (base << PAGE_BITS);
-                ASSERT_MSG(page_table.pointers[base],
+                ASSERT_MSG(page_table.pointers[base].Pointer(),
                           "memory mapping base yield a nullptr within the table");
                base += 1;
@@ -646,21 +593,13 @@ struct Memory::Impl {
    template <typename T>
    T Read(const VAddr vaddr) {
        // Avoid adding any extra logic to this fast-path block
-        if (const u8* const pointer = current_page_table->pointers[vaddr >> PAGE_BITS]) {
+        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
+        if (const u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
            T value;
            std::memcpy(&value, &pointer[vaddr], sizeof(T));
            return value;
        }
+        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        // Otherwise, we need to grab the page with a lock, in case it is currently being modified
-        const auto entry = SafePageEntry(vaddr >> PAGE_BITS);
-        if (entry.pointer) {
-            T value;
-            std::memcpy(&value, &entry.pointer[vaddr], sizeof(T));
-            return value;
-        }
-        switch (entry.attribute) {
        case Common::PageType::Unmapped:
            LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:08X}", sizeof(T) * 8, vaddr);
            return 0;
@@ -692,20 +631,12 @@ struct Memory::Impl {
    template <typename T>
    void Write(const VAddr vaddr, const T data) {
        // Avoid adding any extra logic to this fast-path block
-        if (u8* const pointer = current_page_table->pointers[vaddr >> PAGE_BITS]) {
+        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
+        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
            std::memcpy(&pointer[vaddr], &data, sizeof(T));
            return;
        }
+        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        // Otherwise, we need to grab the page with a lock, in case it is currently being modified
-        const auto entry = SafePageEntry(vaddr >> PAGE_BITS);
-        if (entry.pointer) {
-            // Memory was mapped, we are done
-            std::memcpy(&entry.pointer[vaddr], &data, sizeof(T));
-            return;
-        }
-        switch (entry.attribute) {
        case Common::PageType::Unmapped:
            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
                      static_cast<u32>(data), vaddr);
@@ -726,15 +657,13 @@ struct Memory::Impl {
    template <typename T>
    bool WriteExclusive(const VAddr vaddr, const T data, const T expected) {
-        u8* page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS];
+        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (page_pointer != nullptr) {
+        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
            // NOTE: Avoid adding any extra logic to this fast-path block
-            auto* pointer = reinterpret_cast<volatile T*>(&page_pointer[vaddr]);
+            const auto volatile_pointer = reinterpret_cast<volatile T*>(&pointer[vaddr]);
-            return Common::AtomicCompareAndSwap(pointer, data, expected);
+            return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);
        }
+        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        const Common::PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
-        switch (type) {
        case Common::PageType::Unmapped:
            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}", sizeof(data) * 8,
                      static_cast<u32>(data), vaddr);
@@ -755,15 +684,13 @@ struct Memory::Impl {
    }
    bool WriteExclusive128(const VAddr vaddr, const u128 data, const u128 expected) {
-        u8* const page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS];
+        const uintptr_t raw_pointer = current_page_table->pointers[vaddr >> PAGE_BITS].Raw();
-        if (page_pointer != nullptr) {
+        if (u8* const pointer = Common::PageTable::PageInfo::ExtractPointer(raw_pointer)) {
            // NOTE: Avoid adding any extra logic to this fast-path block
-            auto* pointer = reinterpret_cast<volatile u64*>(&page_pointer[vaddr]);
+            const auto volatile_pointer = reinterpret_cast<volatile u64*>(&pointer[vaddr]);
-            return Common::AtomicCompareAndSwap(pointer, data, expected);
+            return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);
        }
+        switch (Common::PageTable::PageInfo::ExtractType(raw_pointer)) {
-        const Common::PageType type = current_page_table->attributes[vaddr >> PAGE_BITS];
-        switch (type) {
        case Common::PageType::Unmapped:
            LOG_ERROR(HW_Memory, "Unmapped Write{} 0x{:08X} @ 0x{:016X}{:016X}", sizeof(data) * 8,
                      static_cast<u64>(data[1]), static_cast<u64>(data[0]), vaddr);
@@ -783,7 +710,6 @@ struct Memory::Impl {
        return true;
    }
-    mutable std::mutex rasterizer_cache_guard;
    Common::PageTable* current_page_table = nullptr;
    Core::System& system;
 };
@@ -799,25 +725,10 @@ void Memory::MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size
    impl->MapMemoryRegion(page_table, base, size, target);
 }
-void Memory::MapIoRegion(Common::PageTable& page_table, VAddr base, u64 size,
-                         Common::MemoryHookPointer mmio_handler) {
-    impl->MapIoRegion(page_table, base, size, std::move(mmio_handler));
-}
 void Memory::UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size) {
    impl->UnmapRegion(page_table, base, size);
 }
-void Memory::AddDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
-                          Common::MemoryHookPointer hook) {
-    impl->AddDebugHook(page_table, base, size, std::move(hook));
-}
-void Memory::RemoveDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
-                             Common::MemoryHookPointer hook) {
-    impl->RemoveDebugHook(page_table, base, size, std::move(hook));
-}
 bool Memory::IsValidVirtualAddress(const Kernel::Process& process, const VAddr vaddr) const {
    return impl->IsValidVirtualAddress(process, vaddr);
 }
diff --git a/src/core/memory.h b/src/core/memory.h
index 4a1cc63f4..705ebb23d 100644
--- a/src/core/memory.h
+++ b/src/core/memory.h
@@ -8,7 +8,6 @@
 #include <memory>
 #include <string>
 #include "common/common_types.h"
-#include "common/memory_hook.h"
 namespace Common {
 struct PageTable;
@@ -78,17 +77,6 @@ public:
    void MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size, PAddr target);
    /**
-     * Maps a region of the emulated process address space as a IO region.
-     *
-     * @param page_table   The page table of the emulated process.
-     * @param base         The address to start mapping at. Must be page-aligned.
-     * @param size         The amount of bytes to map. Must be page-aligned.
-     * @param mmio_handler The handler that backs the mapping.
-     */
-    void MapIoRegion(Common::PageTable& page_table, VAddr base, u64 size,
-                     Common::MemoryHookPointer mmio_handler);
-    /**
     * Unmaps a region of the emulated process address space.
     *
     * @param page_table The page table of the emulated process.
@@ -98,28 +86,6 @@ public:
    void UnmapRegion(Common::PageTable& page_table, VAddr base, u64 size);
    /**
-     * Adds a memory hook to intercept reads and writes to given region of memory.
-     *
-     * @param page_table The page table of the emulated process
-     * @param base       The starting address to apply the hook to.
-     * @param size       The size of the memory region to apply the hook to, in bytes.
-     * @param hook       The hook to apply to the region of memory.
-     */
-    void AddDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
-                      Common::MemoryHookPointer hook);
-    /**
-     * Removes a memory hook from a given range of memory.
-     *
-     * @param page_table The page table of the emulated process.
-     * @param base       The starting address to remove the hook from.
-     * @param size       The size of the memory region to remove the hook from, in bytes.
-     * @param hook       The hook to remove from the specified region of memory.
-     */
-    void RemoveDebugHook(Common::PageTable& page_table, VAddr base, u64 size,
-                         Common::MemoryHookPointer hook);
-    /**
     * Checks whether or not the supplied address is a valid virtual
     * address for the given process.
     *
diff --git a/src/core/settings.cpp b/src/core/settings.cpp
index 47d9ecf9a..39306509a 100644
--- a/src/core/settings.cpp
+++ b/src/core/settings.cpp
@@ -148,9 +148,4 @@ void RestoreGlobalState(bool is_powered_on) {
    values.motion_enabled.SetGlobal(true);
 }
-void Sanitize() {
-    values.use_asynchronous_gpu_emulation.SetValue(
-        values.use_asynchronous_gpu_emulation.GetValue() || values.use_multi_core.GetValue());
-}
 } // namespace Settings
diff --git a/src/core/settings.h b/src/core/settings.h
index d5f8d2b7e..a324530bd 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -131,6 +131,7 @@ struct Values {
    bool cpuopt_unsafe_unfuse_fma;
    bool cpuopt_unsafe_reduce_fp_error;
+    bool cpuopt_unsafe_inaccurate_nan;
    // Renderer
    Setting<RendererBackend> renderer_backend;
@@ -221,7 +222,7 @@ struct Values {
    bool disable_macro_jit;
    bool extended_logging;
-    // Misceallaneous
+    // Miscellaneous
    std::string log_filter;
    bool use_dev_keys;
@@ -257,7 +258,4 @@ void LogSettings();
 // Restore the global state of all applicable settings in the Values struct
 void RestoreGlobalState(bool is_powered_on);
-// Fixes settings that are known to cause issues with the emulator
-void Sanitize();
 } // namespace Settings
diff --git a/src/input_common/gcadapter/gc_adapter.h b/src/input_common/gcadapter/gc_adapter.h
index f1256c9da..7a6c545bd 100644
--- a/src/input_common/gcadapter/gc_adapter.h
+++ b/src/input_common/gcadapter/gc_adapter.h
@@ -120,17 +120,17 @@ private:
    /// For use in initialization, querying devices to find the adapter
    void Setup();
-    /// Resets status of all GC controller devices to a disconected state
+    /// Resets status of all GC controller devices to a disconnected state
    void ResetDevices();
-    /// Resets status of device connected to a disconected state
+    /// Resets status of device connected to a disconnected state
    void ResetDevice(std::size_t port);
    /// Returns true if we successfully gain access to GC Adapter
    bool CheckDeviceAccess();
    /// Captures GC Adapter endpoint address
-    /// Returns true if the endpoind was set correctly
+    /// Returns true if the endpoint was set correctly
    bool GetGCEndpoint(libusb_device* device);
    /// For shutting down, clear all data, join all threads, release usb
diff --git a/src/input_common/motion_input.cpp b/src/input_common/motion_input.cpp
index f77ba535d..6a65f175e 100644
--- a/src/input_common/motion_input.cpp
+++ b/src/input_common/motion_input.cpp
@@ -129,7 +129,7 @@ void MotionInput::UpdateOrientation(u64 elapsed_time) {
            rad_gyro += ki * integral_error;
            rad_gyro += kd * derivative_error;
        } else {
-            // Give more weight to acelerometer values to compensate for the lack of gyro
+            // Give more weight to accelerometer values to compensate for the lack of gyro
            rad_gyro += 35.0f * kp * real_error;
            rad_gyro += 10.0f * ki * integral_error;
            rad_gyro += 10.0f * kd * derivative_error;
diff --git a/src/input_common/mouse/mouse_input.h b/src/input_common/mouse/mouse_input.h
index 65e64bee7..58803c1bf 100644
--- a/src/input_common/mouse/mouse_input.h
+++ b/src/input_common/mouse/mouse_input.h
@@ -20,7 +20,7 @@ enum class MouseButton {
    Left,
    Wheel,
    Right,
-    Foward,
+    Forward,
    Backward,
    Undefined,
 };
diff --git a/src/input_common/sdl/sdl_impl.cpp b/src/input_common/sdl/sdl_impl.cpp
index 0b531f698..d32eb732a 100644
--- a/src/input_common/sdl/sdl_impl.cpp
+++ b/src/input_common/sdl/sdl_impl.cpp
@@ -1030,11 +1030,44 @@ public:
        }
        return {};
    }
-    [[nodiscard]] std::optional<Common::ParamPackage> FromEvent(const SDL_Event& event) const {
+    [[nodiscard]] std::optional<Common::ParamPackage> FromEvent(SDL_Event& event) {
        switch (event.type) {
        case SDL_JOYAXISMOTION:
-            if (std::abs(event.jaxis.value / 32767.0) < 0.5) {
+            if (!axis_memory.count(event.jaxis.which) ||
+                !axis_memory[event.jaxis.which].count(event.jaxis.axis)) {
+                axis_memory[event.jaxis.which][event.jaxis.axis] = event.jaxis.value;
+                axis_event_count[event.jaxis.which][event.jaxis.axis] = 1;
                break;
+            } else {
+                axis_event_count[event.jaxis.which][event.jaxis.axis]++;
+                // The joystick and axis exist in our map if we take this branch, so no checks
+                // needed
+                if (std::abs(
+                        (event.jaxis.value - axis_memory[event.jaxis.which][event.jaxis.axis]) /
+                        32767.0) < 0.5) {
+                    break;
+                } else {
+                    if (axis_event_count[event.jaxis.which][event.jaxis.axis] == 2 &&
+                        IsAxisAtPole(event.jaxis.value) &&
+                        IsAxisAtPole(axis_memory[event.jaxis.which][event.jaxis.axis])) {
+                        // If we have exactly two events and both are near a pole, this is
+                        // likely a digital input masquerading as an analog axis; Instead of
+                        // trying to look at the direction the axis travelled, assume the first
+                        // event was press and the second was release; This should handle most
+                        // digital axes while deferring to the direction of travel for analog
+                        // axes
+                        event.jaxis.value = static_cast<Sint16>(
+                            std::copysign(32767, axis_memory[event.jaxis.which][event.jaxis.axis]));
+                    } else {
+                        // There are more than two events, so this is likely a true analog axis,
+                        // check the direction it travelled
+                        event.jaxis.value = static_cast<Sint16>(std::copysign(
+                            32767,
+                            event.jaxis.value - axis_memory[event.jaxis.which][event.jaxis.axis]));
+                    }
+                    axis_memory.clear();
+                    axis_event_count.clear();
+                }
            }
            [[fallthrough]];
        case SDL_JOYBUTTONUP:
@@ -1043,6 +1076,16 @@ public:
        }
        return std::nullopt;
    }
+private:
+    // Determine whether an axis value is close to an extreme or center
+    // Some controllers have a digital D-Pad as a pair of analog sticks, with 3 possible values per
+    // axis, which is why the center must be considered a pole
+    bool IsAxisAtPole(int16_t value) const {
+        return std::abs(value) >= 32767 || std::abs(value) < 327;
+    }
+    std::unordered_map<SDL_JoystickID, std::unordered_map<uint8_t, int16_t>> axis_memory;
+    std::unordered_map<SDL_JoystickID, std::unordered_map<uint8_t, uint32_t>> axis_event_count;
 };
 class SDLMotionPoller final : public SDLPoller {
diff --git a/src/input_common/udp/udp.cpp b/src/input_common/udp/udp.cpp
index 8686a059c..c5da27a38 100644
--- a/src/input_common/udp/udp.cpp
+++ b/src/input_common/udp/udp.cpp
@@ -28,14 +28,14 @@ private:
    mutable std::mutex mutex;
 };
-/// A motion device factory that creates motion devices from JC Adapter
+/// A motion device factory that creates motion devices from a UDP client
 UDPMotionFactory::UDPMotionFactory(std::shared_ptr<CemuhookUDP::Client> client_)
    : client(std::move(client_)) {}
 /**
 * Creates motion device
 * @param params contains parameters for creating the device:
- *     - "port": the nth jcpad on the adapter
+ *     - "port": the UDP port number
 */
 std::unique_ptr<Input::MotionDevice> UDPMotionFactory::Create(const Common::ParamPackage& params) {
    auto ip = params.Get("ip", "127.0.0.1");
@@ -90,14 +90,14 @@ private:
    mutable std::mutex mutex;
 };
-/// A motion device factory that creates motion devices from JC Adapter
+/// A motion device factory that creates motion devices from a UDP client
 UDPTouchFactory::UDPTouchFactory(std::shared_ptr<CemuhookUDP::Client> client_)
    : client(std::move(client_)) {}
 /**
 * Creates motion device
 * @param params contains parameters for creating the device:
- *     - "port": the nth jcpad on the adapter
+ *     - "port": the UDP port number
 */
 std::unique_ptr<Input::TouchDevice> UDPTouchFactory::Create(const Common::ParamPackage& params) {
    auto ip = params.Get("ip", "127.0.0.1");
diff --git a/src/tests/CMakeLists.txt b/src/tests/CMakeLists.txt
index d80b0b688..8a606b448 100644
--- a/src/tests/CMakeLists.txt
+++ b/src/tests/CMakeLists.txt
@@ -4,8 +4,6 @@ add_executable(tests
    common/fibers.cpp
    common/param_package.cpp
    common/ring_buffer.cpp
-    core/arm/arm_test_common.cpp
-    core/arm/arm_test_common.h
    core/core_timing.cpp
    tests.cpp
 )
diff --git a/src/tests/common/fibers.cpp b/src/tests/common/fibers.cpp
index 4757dd2b4..d94492fc6 100644
--- a/src/tests/common/fibers.cpp
+++ b/src/tests/common/fibers.cpp
@@ -207,7 +207,7 @@ static void ThreadStart2_2(u32 id, TestControl2& test_control) {
 }
 /** This test checks for fiber thread exchange configuration and validates that fibers are
- *  that a fiber has been succesfully transfered from one thread to another and that the TLS
+ *  that a fiber has been successfully transferred from one thread to another and that the TLS
 *  region of the thread is kept while changing fibers.
 */
 TEST_CASE("Fibers::InterExchange", "[common]") {
@@ -299,7 +299,7 @@ static void ThreadStart3(u32 id, TestControl3& test_control) {
 }
 /** This test checks for one two threads racing for starting the same fiber.
- *  It checks execution occured in an ordered manner and by no time there were
+ *  It checks execution occurred in an ordered manner and by no time there were
 *  two contexts at the same time.
 */
 TEST_CASE("Fibers::StartRace", "[common]") {
diff --git a/src/tests/core/arm/arm_test_common.cpp b/src/tests/core/arm/arm_test_common.cpp
deleted file mode 100644
index e54674d11..000000000
--- a/src/tests/core/arm/arm_test_common.cpp
+++ /dev/null
@@ -1,145 +0,0 @@
-// Copyright 2016 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#include <algorithm>
-#include "common/page_table.h"
-#include "core/core.h"
-#include "core/hle/kernel/memory/page_table.h"
-#include "core/hle/kernel/process.h"
-#include "core/memory.h"
-#include "tests/core/arm/arm_test_common.h"
-namespace ArmTests {
-TestEnvironment::TestEnvironment(bool mutable_memory_)
-    : mutable_memory(mutable_memory_),
-      test_memory(std::make_shared<TestMemory>(this)), kernel{Core::System::GetInstance()} {
-    auto& system = Core::System::GetInstance();
-    auto process = Kernel::Process::Create(system, "", Kernel::Process::ProcessType::Userland);
-    page_table = &process->PageTable().PageTableImpl();
-    system.Memory().MapIoRegion(*page_table, 0x00000000, 0x80000000, test_memory);
-    system.Memory().MapIoRegion(*page_table, 0x80000000, 0x80000000, test_memory);
-    kernel.MakeCurrentProcess(process.get());
-}
-TestEnvironment::~TestEnvironment() {
-    auto& system = Core::System::GetInstance();
-    system.Memory().UnmapRegion(*page_table, 0x80000000, 0x80000000);
-    system.Memory().UnmapRegion(*page_table, 0x00000000, 0x80000000);
-}
-void TestEnvironment::SetMemory64(VAddr vaddr, u64 value) {
-    SetMemory32(vaddr + 0, static_cast<u32>(value));
-    SetMemory32(vaddr + 4, static_cast<u32>(value >> 32));
-}
-void TestEnvironment::SetMemory32(VAddr vaddr, u32 value) {
-    SetMemory16(vaddr + 0, static_cast<u16>(value));
-    SetMemory16(vaddr + 2, static_cast<u16>(value >> 16));
-}
-void TestEnvironment::SetMemory16(VAddr vaddr, u16 value) {
-    SetMemory8(vaddr + 0, static_cast<u8>(value));
-    SetMemory8(vaddr + 1, static_cast<u8>(value >> 8));
-}
-void TestEnvironment::SetMemory8(VAddr vaddr, u8 value) {
-    test_memory->data[vaddr] = value;
-}
-std::vector<WriteRecord> TestEnvironment::GetWriteRecords() const {
-    return write_records;
-}
-void TestEnvironment::ClearWriteRecords() {
-    write_records.clear();
-}
-TestEnvironment::TestMemory::~TestMemory() {}
-std::optional<bool> TestEnvironment::TestMemory::IsValidAddress(VAddr addr) {
-    return true;
-}
-std::optional<u8> TestEnvironment::TestMemory::Read8(VAddr addr) {
-    const auto iter = data.find(addr);
-    if (iter == data.end()) {
-        // Some arbitrary data
-        return static_cast<u8>(addr);
-    }
-    return iter->second;
-}
-std::optional<u16> TestEnvironment::TestMemory::Read16(VAddr addr) {
-    return *Read8(addr) | static_cast<u16>(*Read8(addr + 1)) << 8;
-}
-std::optional<u32> TestEnvironment::TestMemory::Read32(VAddr addr) {
-    return *Read16(addr) | static_cast<u32>(*Read16(addr + 2)) << 16;
-}
-std::optional<u64> TestEnvironment::TestMemory::Read64(VAddr addr) {
-    return *Read32(addr) | static_cast<u64>(*Read32(addr + 4)) << 32;
-}
-bool TestEnvironment::TestMemory::ReadBlock(VAddr src_addr, void* dest_buffer, std::size_t size) {
-    VAddr addr = src_addr;
-    u8* data = static_cast<u8*>(dest_buffer);
-    for (std::size_t i = 0; i < size; i++, addr++, data++) {
-        *data = *Read8(addr);
-    }
-    return true;
-}
-bool TestEnvironment::TestMemory::Write8(VAddr addr, u8 data) {
-    env->write_records.emplace_back(8, addr, data);
-    if (env->mutable_memory)
-        env->SetMemory8(addr, data);
-    return true;
-}
-bool TestEnvironment::TestMemory::Write16(VAddr addr, u16 data) {
-    env->write_records.emplace_back(16, addr, data);
-    if (env->mutable_memory)
-        env->SetMemory16(addr, data);
-    return true;
-}
-bool TestEnvironment::TestMemory::Write32(VAddr addr, u32 data) {
-    env->write_records.emplace_back(32, addr, data);
-    if (env->mutable_memory)
-        env->SetMemory32(addr, data);
-    return true;
-}
-bool TestEnvironment::TestMemory::Write64(VAddr addr, u64 data) {
-    env->write_records.emplace_back(64, addr, data);
-    if (env->mutable_memory)
-        env->SetMemory64(addr, data);
-    return true;
-}
-bool TestEnvironment::TestMemory::WriteBlock(VAddr dest_addr, const void* src_buffer,
-                                             std::size_t size) {
-    VAddr addr = dest_addr;
-    const u8* data = static_cast<const u8*>(src_buffer);
-    for (std::size_t i = 0; i < size; i++, addr++, data++) {
-        env->write_records.emplace_back(8, addr, *data);
-        if (env->mutable_memory)
-            env->SetMemory8(addr, *data);
-    }
-    return true;
-}
-} // namespace ArmTests
diff --git a/src/tests/core/arm/arm_test_common.h b/src/tests/core/arm/arm_test_common.h
deleted file mode 100644
index d145dbfcc..000000000
--- a/src/tests/core/arm/arm_test_common.h
+++ /dev/null
@@ -1,93 +0,0 @@
-// Copyright 2016 Citra Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#pragma once
-#include <tuple>
-#include <unordered_map>
-#include <vector>
-#include "common/common_types.h"
-#include "common/memory_hook.h"
-#include "core/hle/kernel/kernel.h"
-namespace Common {
-struct PageTable;
-}
-namespace ArmTests {
-struct WriteRecord {
-    WriteRecord(std::size_t size, VAddr addr, u64 data) : size(size), addr(addr), data(data) {}
-    std::size_t size;
-    VAddr addr;
-    u64 data;
-    bool operator==(const WriteRecord& o) const {
-        return std::tie(size, addr, data) == std::tie(o.size, o.addr, o.data);
-    }
-};
-class TestEnvironment final {
-public:
-    /*
-     * Inititalise test environment
-     * @param mutable_memory If false, writes to memory can never be read back.
-     *                       (Memory is immutable.)
-     */
-    explicit TestEnvironment(bool mutable_memory = false);
-    /// Shutdown test environment
-    ~TestEnvironment();
-    /// Sets value at memory location vaddr.
-    void SetMemory8(VAddr vaddr, u8 value);
-    void SetMemory16(VAddr vaddr, u16 value);
-    void SetMemory32(VAddr vaddr, u32 value);
-    void SetMemory64(VAddr vaddr, u64 value);
-    /**
-     * Whenever Memory::Write{8,16,32,64} is called within the test environment,
-     * a new write-record is made.
-     * @returns A vector of write records made since they were last cleared.
-     */
-    std::vector<WriteRecord> GetWriteRecords() const;
-    /// Empties the internal write-record store.
-    void ClearWriteRecords();
-private:
-    friend struct TestMemory;
-    struct TestMemory final : Common::MemoryHook {
-        explicit TestMemory(TestEnvironment* env_) : env(env_) {}
-        TestEnvironment* env;
-        ~TestMemory() override;
-        std::optional<bool> IsValidAddress(VAddr addr) override;
-        std::optional<u8> Read8(VAddr addr) override;
-        std::optional<u16> Read16(VAddr addr) override;
-        std::optional<u32> Read32(VAddr addr) override;
-        std::optional<u64> Read64(VAddr addr) override;
-        bool ReadBlock(VAddr src_addr, void* dest_buffer, std::size_t size) override;
-        bool Write8(VAddr addr, u8 data) override;
-        bool Write16(VAddr addr, u16 data) override;
-        bool Write32(VAddr addr, u32 data) override;
-        bool Write64(VAddr addr, u64 data) override;
-        bool WriteBlock(VAddr dest_addr, const void* src_buffer, std::size_t size) override;
-        std::unordered_map<VAddr, u8> data;
-    };
-    bool mutable_memory;
-    std::shared_ptr<TestMemory> test_memory;
-    std::vector<WriteRecord> write_records;
-    Common::PageTable* page_table = nullptr;
-    Kernel::KernelCore kernel;
-};
-} // namespace ArmTests
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index f977cf12b..4bd48f706 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -48,6 +48,7 @@ add_library(video_core STATIC
    engines/shader_bytecode.h
    engines/shader_header.h
    engines/shader_type.h
+    framebuffer_config.h
    macro/macro.cpp
    macro/macro.h
    macro/macro_hle.cpp
@@ -59,10 +60,6 @@ add_library(video_core STATIC
    fence_manager.h
    gpu.cpp
    gpu.h
-    gpu_asynch.cpp
-    gpu_asynch.h
-    gpu_synch.cpp
-    gpu_synch.h
    gpu_thread.cpp
    gpu_thread.h
    guest_driver.cpp
diff --git a/src/video_core/command_classes/vic.cpp b/src/video_core/command_classes/vic.cpp
index aa8c9f9de..55e632346 100644
--- a/src/video_core/command_classes/vic.cpp
+++ b/src/video_core/command_classes/vic.cpp
@@ -53,7 +53,7 @@ void Vic::ProcessMethod(Method method, const std::vector<u32>& arguments) {
 void Vic::Execute() {
    if (output_surface_luma_address == 0) {
-        LOG_ERROR(Service_NVDRV, "VIC Luma address not set. Recieved 0x{:X}",
+        LOG_ERROR(Service_NVDRV, "VIC Luma address not set. Received 0x{:X}",
                  vic_state.output_surface.luma_offset);
        return;
    }
diff --git a/src/video_core/framebuffer_config.h b/src/video_core/framebuffer_config.h
new file mode 100644
index 000000000..b86c3a757
--- /dev/null
+++ b/src/video_core/framebuffer_config.h
@@ -0,0 +1,31 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+#pragma once
+namespace Tegra {
+/**
+ * Struct describing framebuffer configuration
+ */
+struct FramebufferConfig {
+    enum class PixelFormat : u32 {
+        A8B8G8R8_UNORM = 1,
+        RGB565_UNORM = 4,
+        B8G8R8A8_UNORM = 5,
+    };
+    VAddr address{};
+    u32 offset{};
+    u32 width{};
+    u32 height{};
+    u32 stride{};
+    PixelFormat pixel_format{};
+    using TransformFlags = Service::NVFlinger::BufferQueue::BufferTransformFlags;
+    TransformFlags transform_flags{};
+    Common::Rectangle<int> crop_rect;
+};
+} // namespace Tegra
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index e2512a7f2..6ab06775f 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -10,6 +10,7 @@
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/frontend/emu_window.h"
+#include "core/hardware_interrupt_manager.h"
 #include "core/memory.h"
 #include "core/settings.h"
 #include "video_core/engines/fermi_2d.h"
@@ -36,7 +37,8 @@ GPU::GPU(Core::System& system_, bool is_async_, bool use_nvdec_)
      kepler_compute{std::make_unique<Engines::KeplerCompute>(system, *memory_manager)},
      maxwell_dma{std::make_unique<Engines::MaxwellDMA>(system, *memory_manager)},
      kepler_memory{std::make_unique<Engines::KeplerMemory>(system, *memory_manager)},
-      shader_notify{std::make_unique<VideoCore::ShaderNotify>()}, is_async{is_async_} {}
+      shader_notify{std::make_unique<VideoCore::ShaderNotify>()}, is_async{is_async_},
+      gpu_thread{system_, is_async_} {}
 GPU::~GPU() = default;
@@ -198,10 +200,6 @@ void GPU::SyncGuestHost() {
    renderer->Rasterizer().SyncGuestHost();
 }
-void GPU::OnCommandListEnd() {
-    renderer->Rasterizer().ReleaseFences();
-}
 enum class GpuSemaphoreOperation {
    AcquireEqual = 0x1,
    WriteLong = 0x2,
@@ -461,4 +459,75 @@ void GPU::ProcessSemaphoreAcquire() {
    }
 }
+void GPU::Start() {
+    gpu_thread.StartThread(*renderer, renderer->Context(), *dma_pusher, *cdma_pusher);
+    cpu_context = renderer->GetRenderWindow().CreateSharedContext();
+    cpu_context->MakeCurrent();
+}
+void GPU::ObtainContext() {
+    cpu_context->MakeCurrent();
+}
+void GPU::ReleaseContext() {
+    cpu_context->DoneCurrent();
+}
+void GPU::PushGPUEntries(Tegra::CommandList&& entries) {
+    gpu_thread.SubmitList(std::move(entries));
+}
+void GPU::PushCommandBuffer(Tegra::ChCommandHeaderList& entries) {
+    if (!use_nvdec) {
+        return;
+    }
+    // This condition fires when a video stream ends, clear all intermediary data
+    if (entries[0].raw == 0xDEADB33F) {
+        cdma_pusher.reset();
+        return;
+    }
+    if (!cdma_pusher) {
+        cdma_pusher = std::make_unique<Tegra::CDmaPusher>(*this);
+    }
+    // SubmitCommandBuffer would make the nvdec operations async, this is not currently working
+    // TODO(ameerj): RE proper async nvdec operation
+    // gpu_thread.SubmitCommandBuffer(std::move(entries));
+    cdma_pusher->Push(std::move(entries));
+    cdma_pusher->DispatchCalls();
+}
+void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
+    gpu_thread.SwapBuffers(framebuffer);
+}
+void GPU::FlushRegion(VAddr addr, u64 size) {
+    gpu_thread.FlushRegion(addr, size);
+}
+void GPU::InvalidateRegion(VAddr addr, u64 size) {
+    gpu_thread.InvalidateRegion(addr, size);
+}
+void GPU::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    gpu_thread.FlushAndInvalidateRegion(addr, size);
+}
+void GPU::TriggerCpuInterrupt(const u32 syncpoint_id, const u32 value) const {
+    auto& interrupt_manager = system.InterruptManager();
+    interrupt_manager.GPUInterruptSyncpt(syncpoint_id, value);
+}
+void GPU::WaitIdle() const {
+    gpu_thread.WaitIdle();
+}
+void GPU::OnCommandListEnd() {
+    if (is_async) {
+        // This command only applies to asynchronous GPU mode
+        gpu_thread.OnCommandListEnd();
+    }
+}
 } // namespace Tegra
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 660641d04..d81e38680 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -15,6 +15,8 @@
 #include "core/hle/service/nvflinger/buffer_queue.h"
 #include "video_core/cdma_pusher.h"
 #include "video_core/dma_pusher.h"
+#include "video_core/framebuffer_config.h"
+#include "video_core/gpu_thread.h"
 using CacheAddr = std::uintptr_t;
 [[nodiscard]] inline CacheAddr ToCacheAddr(const void* host_ptr) {
@@ -101,28 +103,6 @@ enum class DepthFormat : u32 {
 struct CommandListHeader;
 class DebugContext;
-/**
- * Struct describing framebuffer configuration
- */
-struct FramebufferConfig {
-    enum class PixelFormat : u32 {
-        A8B8G8R8_UNORM = 1,
-        RGB565_UNORM = 4,
-        B8G8R8A8_UNORM = 5,
-    };
-    VAddr address;
-    u32 offset;
-    u32 width;
-    u32 height;
-    u32 stride;
-    PixelFormat pixel_format;
-    using TransformFlags = Service::NVFlinger::BufferQueue::BufferTransformFlags;
-    TransformFlags transform_flags;
-    Common::Rectangle<int> crop_rect;
-};
 namespace Engines {
 class Fermi2D;
 class Maxwell3D;
@@ -141,7 +121,7 @@ enum class EngineID {
 class MemoryManager;
-class GPU {
+class GPU final {
 public:
    struct MethodCall {
        u32 method{};
@@ -159,7 +139,7 @@ public:
    };
    explicit GPU(Core::System& system_, bool is_async_, bool use_nvdec_);
-    virtual ~GPU();
+    ~GPU();
    /// Binds a renderer to the GPU.
    void BindRenderer(std::unique_ptr<VideoCore::RendererBase> renderer);
@@ -176,7 +156,7 @@ public:
    /// Synchronizes CPU writes with Host GPU memory.
    void SyncGuestHost();
    /// Signal the ending of command list.
-    virtual void OnCommandListEnd();
+    void OnCommandListEnd();
    /// Request a host GPU memory flush from the CPU.
    [[nodiscard]] u64 RequestFlush(VAddr addr, std::size_t size);
@@ -240,7 +220,7 @@ public:
    }
    // Waits for the GPU to finish working
-    virtual void WaitIdle() const = 0;
+    void WaitIdle() const;
    /// Allows the CPU/NvFlinger to wait on the GPU before presenting a frame.
    void WaitFence(u32 syncpoint_id, u32 value);
@@ -330,34 +310,34 @@ public:
    /// Performs any additional setup necessary in order to begin GPU emulation.
    /// This can be used to launch any necessary threads and register any necessary
    /// core timing events.
-    virtual void Start() = 0;
+    void Start();
    /// Obtain the CPU Context
-    virtual void ObtainContext() = 0;
+    void ObtainContext();
    /// Release the CPU Context
-    virtual void ReleaseContext() = 0;
+    void ReleaseContext();
    /// Push GPU command entries to be processed
-    virtual void PushGPUEntries(Tegra::CommandList&& entries) = 0;
+    void PushGPUEntries(Tegra::CommandList&& entries);
    /// Push GPU command buffer entries to be processed
-    virtual void PushCommandBuffer(Tegra::ChCommandHeaderList& entries) = 0;
+    void PushCommandBuffer(Tegra::ChCommandHeaderList& entries);
    /// Swap buffers (render frame)
-    virtual void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) = 0;
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
    /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    virtual void FlushRegion(VAddr addr, u64 size) = 0;
+    void FlushRegion(VAddr addr, u64 size);
    /// Notify rasterizer that any caches of the specified region should be invalidated
-    virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
+    void InvalidateRegion(VAddr addr, u64 size);
    /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
-    virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
+    void FlushAndInvalidateRegion(VAddr addr, u64 size);
 protected:
-    virtual void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const = 0;
+    void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const;
 private:
    void ProcessBindMethod(const MethodCall& method_call);
@@ -427,6 +407,9 @@ private:
    std::mutex flush_request_mutex;
    const bool is_async;
+    VideoCommon::GPUThread::ThreadManager gpu_thread;
+    std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context;
 };
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp
deleted file mode 100644
index 6cc091ecd..000000000
--- a/src/video_core/gpu_asynch.cpp
+++ /dev/null
@@ -1,86 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#include "core/core.h"
-#include "core/hardware_interrupt_manager.h"
-#include "video_core/gpu_asynch.h"
-#include "video_core/gpu_thread.h"
-#include "video_core/renderer_base.h"
-namespace VideoCommon {
-GPUAsynch::GPUAsynch(Core::System& system_, bool use_nvdec_)
-    : GPU{system_, true, use_nvdec_}, gpu_thread{system_} {}
-GPUAsynch::~GPUAsynch() = default;
-void GPUAsynch::Start() {
-    gpu_thread.StartThread(*renderer, renderer->Context(), *dma_pusher, *cdma_pusher);
-    cpu_context = renderer->GetRenderWindow().CreateSharedContext();
-    cpu_context->MakeCurrent();
-}
-void GPUAsynch::ObtainContext() {
-    cpu_context->MakeCurrent();
-}
-void GPUAsynch::ReleaseContext() {
-    cpu_context->DoneCurrent();
-}
-void GPUAsynch::PushGPUEntries(Tegra::CommandList&& entries) {
-    gpu_thread.SubmitList(std::move(entries));
-}
-void GPUAsynch::PushCommandBuffer(Tegra::ChCommandHeaderList& entries) {
-    if (!use_nvdec) {
-        return;
-    }
-    // This condition fires when a video stream ends, clear all intermediary data
-    if (entries[0].raw == 0xDEADB33F) {
-        cdma_pusher.reset();
-        return;
-    }
-    if (!cdma_pusher) {
-        cdma_pusher = std::make_unique<Tegra::CDmaPusher>(*this);
-    }
-    // SubmitCommandBuffer would make the nvdec operations async, this is not currently working
-    // TODO(ameerj): RE proper async nvdec operation
-    // gpu_thread.SubmitCommandBuffer(std::move(entries));
-    cdma_pusher->Push(std::move(entries));
-    cdma_pusher->DispatchCalls();
-}
-void GPUAsynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
-    gpu_thread.SwapBuffers(framebuffer);
-}
-void GPUAsynch::FlushRegion(VAddr addr, u64 size) {
-    gpu_thread.FlushRegion(addr, size);
-}
-void GPUAsynch::InvalidateRegion(VAddr addr, u64 size) {
-    gpu_thread.InvalidateRegion(addr, size);
-}
-void GPUAsynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
-    gpu_thread.FlushAndInvalidateRegion(addr, size);
-}
-void GPUAsynch::TriggerCpuInterrupt(const u32 syncpoint_id, const u32 value) const {
-    auto& interrupt_manager = system.InterruptManager();
-    interrupt_manager.GPUInterruptSyncpt(syncpoint_id, value);
-}
-void GPUAsynch::WaitIdle() const {
-    gpu_thread.WaitIdle();
-}
-void GPUAsynch::OnCommandListEnd() {
-    gpu_thread.OnCommandListEnd();
-}
-} // namespace VideoCommon
diff --git a/src/video_core/gpu_asynch.h b/src/video_core/gpu_asynch.h
deleted file mode 100644
index a384113f4..000000000
--- a/src/video_core/gpu_asynch.h
+++ /dev/null
@@ -1,47 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#pragma once
-#include "video_core/gpu.h"
-#include "video_core/gpu_thread.h"
-namespace Core::Frontend {
-class GraphicsContext;
-}
-namespace VideoCore {
-class RendererBase;
-} // namespace VideoCore
-namespace VideoCommon {
-/// Implementation of GPU interface that runs the GPU asynchronously
-class GPUAsynch final : public Tegra::GPU {
-public:
-    explicit GPUAsynch(Core::System& system_, bool use_nvdec_);
-    ~GPUAsynch() override;
-    void Start() override;
-    void ObtainContext() override;
-    void ReleaseContext() override;
-    void PushGPUEntries(Tegra::CommandList&& entries) override;
-    void PushCommandBuffer(Tegra::ChCommandHeaderList& entries) override;
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    void FlushRegion(VAddr addr, u64 size) override;
-    void InvalidateRegion(VAddr addr, u64 size) override;
-    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
-    void WaitIdle() const override;
-    void OnCommandListEnd() override;
-protected:
-    void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const override;
-private:
-    GPUThread::ThreadManager gpu_thread;
-    std::unique_ptr<Core::Frontend::GraphicsContext> cpu_context;
-};
-} // namespace VideoCommon
diff --git a/src/video_core/gpu_synch.cpp b/src/video_core/gpu_synch.cpp
deleted file mode 100644
index 1e9d4b9b2..000000000
--- a/src/video_core/gpu_synch.cpp
+++ /dev/null
@@ -1,61 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#include "video_core/gpu_synch.h"
-#include "video_core/renderer_base.h"
-namespace VideoCommon {
-GPUSynch::GPUSynch(Core::System& system_, bool use_nvdec_) : GPU{system_, false, use_nvdec_} {}
-GPUSynch::~GPUSynch() = default;
-void GPUSynch::Start() {}
-void GPUSynch::ObtainContext() {
-    renderer->Context().MakeCurrent();
-}
-void GPUSynch::ReleaseContext() {
-    renderer->Context().DoneCurrent();
-}
-void GPUSynch::PushGPUEntries(Tegra::CommandList&& entries) {
-    dma_pusher->Push(std::move(entries));
-    dma_pusher->DispatchCalls();
-}
-void GPUSynch::PushCommandBuffer(Tegra::ChCommandHeaderList& entries) {
-    if (!use_nvdec) {
-        return;
-    }
-    // This condition fires when a video stream ends, clears all intermediary data
-    if (entries[0].raw == 0xDEADB33F) {
-        cdma_pusher.reset();
-        return;
-    }
-    if (!cdma_pusher) {
-        cdma_pusher = std::make_unique<Tegra::CDmaPusher>(*this);
-    }
-    cdma_pusher->Push(std::move(entries));
-    cdma_pusher->DispatchCalls();
-}
-void GPUSynch::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
-    renderer->SwapBuffers(framebuffer);
-}
-void GPUSynch::FlushRegion(VAddr addr, u64 size) {
-    renderer->Rasterizer().FlushRegion(addr, size);
-}
-void GPUSynch::InvalidateRegion(VAddr addr, u64 size) {
-    renderer->Rasterizer().InvalidateRegion(addr, size);
-}
-void GPUSynch::FlushAndInvalidateRegion(VAddr addr, u64 size) {
-    renderer->Rasterizer().FlushAndInvalidateRegion(addr, size);
-}
-} // namespace VideoCommon
diff --git a/src/video_core/gpu_synch.h b/src/video_core/gpu_synch.h
deleted file mode 100644
index c5904b8db..000000000
--- a/src/video_core/gpu_synch.h
+++ /dev/null
@@ -1,41 +0,0 @@
-// Copyright 2019 yuzu Emulator Project
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-#pragma once
-#include "video_core/gpu.h"
-namespace Core::Frontend {
-class GraphicsContext;
-}
-namespace VideoCore {
-class RendererBase;
-} // namespace VideoCore
-namespace VideoCommon {
-/// Implementation of GPU interface that runs the GPU synchronously
-class GPUSynch final : public Tegra::GPU {
-public:
-    explicit GPUSynch(Core::System& system_, bool use_nvdec_);
-    ~GPUSynch() override;
-    void Start() override;
-    void ObtainContext() override;
-    void ReleaseContext() override;
-    void PushGPUEntries(Tegra::CommandList&& entries) override;
-    void PushCommandBuffer(Tegra::ChCommandHeaderList& entries) override;
-    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
-    void FlushRegion(VAddr addr, u64 size) override;
-    void InvalidateRegion(VAddr addr, u64 size) override;
-    void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
-    void WaitIdle() const override {}
-protected:
-    void TriggerCpuInterrupt([[maybe_unused]] u32 syncpoint_id,
-                             [[maybe_unused]] u32 value) const override {}
-};
-} // namespace VideoCommon
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index e27218b96..7e490bcc3 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -4,6 +4,7 @@
 #include "common/assert.h"
 #include "common/microprofile.h"
+#include "common/scope_exit.h"
 #include "common/thread.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
@@ -21,6 +22,8 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
                      SynchState& state, Tegra::CDmaPusher& cdma_pusher) {
    std::string name = "yuzu:GPU";
    MicroProfileOnThreadCreate(name.c_str());
+    SCOPE_EXIT({ MicroProfileOnThreadExit(); });
    Common::SetCurrentThreadName(name.c_str());
    Common::SetCurrentThreadPriority(Common::ThreadPriority::High);
    system.RegisterHostThread();
@@ -65,7 +68,8 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
    }
 }
-ThreadManager::ThreadManager(Core::System& system_) : system{system_} {}
+ThreadManager::ThreadManager(Core::System& system_, bool is_async_)
+    : system{system_}, is_async{is_async_} {}
 ThreadManager::~ThreadManager() {
    if (!thread.joinable()) {
@@ -97,19 +101,30 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
 }
 void ThreadManager::FlushRegion(VAddr addr, u64 size) {
-    if (!Settings::IsGPULevelHigh()) {
+    if (!is_async) {
+        // Always flush with synchronous GPU mode
        PushCommand(FlushRegionCommand(addr, size));
        return;
    }
-    if (!Settings::IsGPULevelExtreme()) {
-        return;
+    // Asynchronous GPU mode
-    }
+    switch (Settings::values.gpu_accuracy.GetValue()) {
-    if (system.Renderer().Rasterizer().MustFlushRegion(addr, size)) {
+    case Settings::GPUAccuracy::Normal:
+        PushCommand(FlushRegionCommand(addr, size));
+        break;
+    case Settings::GPUAccuracy::High:
+        // TODO(bunnei): Is this right? Preserving existing behavior for now
+        break;
+    case Settings::GPUAccuracy::Extreme: {
        auto& gpu = system.GPU();
        u64 fence = gpu.RequestFlush(addr, size);
        PushCommand(GPUTickCommand());
        while (fence > gpu.CurrentFlushRequestFence()) {
        }
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unsupported gpu_accuracy {}", Settings::values.gpu_accuracy.GetValue());
    }
 }
@@ -123,7 +138,8 @@ void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
 }
 void ThreadManager::WaitIdle() const {
-    while (state.last_fence > state.signaled_fence.load(std::memory_order_relaxed)) {
+    while (state.last_fence > state.signaled_fence.load(std::memory_order_relaxed) &&
+           system.IsPoweredOn()) {
    }
 }
@@ -134,6 +150,12 @@ void ThreadManager::OnCommandListEnd() {
 u64 ThreadManager::PushCommand(CommandData&& command_data) {
    const u64 fence{++state.last_fence};
    state.queue.Push(CommandDataContainer(std::move(command_data), fence));
+    if (!is_async) {
+        // In synchronous GPU mode, block the caller until the command has executed
+        WaitIdle();
+    }
    return fence;
 }
diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h
index f1c52cd9e..2775629e7 100644
--- a/src/video_core/gpu_thread.h
+++ b/src/video_core/gpu_thread.h
@@ -10,8 +10,9 @@
 #include <optional>
 #include <thread>
 #include <variant>
 #include "common/threadsafe_queue.h"
-#include "video_core/gpu.h"
+#include "video_core/framebuffer_config.h"
 namespace Tegra {
 struct FramebufferConfig;
@@ -25,6 +26,10 @@ class GraphicsContext;
 class System;
 } // namespace Core
+namespace VideoCore {
+class RendererBase;
+} // namespace VideoCore
 namespace VideoCommon::GPUThread {
 /// Command to signal to the GPU thread that processing has ended
@@ -112,7 +117,7 @@ struct SynchState final {
 /// Class used to manage the GPU thread
 class ThreadManager final {
 public:
-    explicit ThreadManager(Core::System& system_);
+    explicit ThreadManager(Core::System& system_, bool is_async_);
    ~ThreadManager();
    /// Creates and starts the GPU thread.
@@ -150,6 +155,7 @@ private:
    Core::System& system;
    std::thread thread;
    std::thread::id thread_id;
+    const bool is_async;
 };
 } // namespace VideoCommon::GPUThread
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index fd55ca8a8..9008530d5 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -484,7 +484,7 @@ VkFormat VKDevice::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFla
 }
 void VKDevice::ReportLoss() const {
-    LOG_CRITICAL(Render_Vulkan, "Device loss occured!");
+    LOG_CRITICAL(Render_Vulkan, "Device loss occurred!");
    // Wait for the log to flush and for Nsight Aftermath to dump the results
    std::this_thread::sleep_for(std::chrono::seconds{15});
diff --git a/src/video_core/video_core.cpp b/src/video_core/video_core.cpp
index 837800bfe..53444e945 100644
--- a/src/video_core/video_core.cpp
+++ b/src/video_core/video_core.cpp
@@ -7,8 +7,6 @@
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/settings.h"
-#include "video_core/gpu_asynch.h"
-#include "video_core/gpu_synch.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
 #include "video_core/renderer_vulkan/renderer_vulkan.h"
@@ -39,13 +37,9 @@ std::unique_ptr<VideoCore::RendererBase> CreateRenderer(
 namespace VideoCore {
 std::unique_ptr<Tegra::GPU> CreateGPU(Core::Frontend::EmuWindow& emu_window, Core::System& system) {
-    std::unique_ptr<Tegra::GPU> gpu;
    const bool use_nvdec = Settings::values.use_nvdec_emulation.GetValue();
-    if (Settings::values.use_asynchronous_gpu_emulation.GetValue()) {
+    std::unique_ptr<Tegra::GPU> gpu = std::make_unique<Tegra::GPU>(
-        gpu = std::make_unique<VideoCommon::GPUAsynch>(system, use_nvdec);
+        system, Settings::values.use_asynchronous_gpu_emulation.GetValue(), use_nvdec);
-    } else {
-        gpu = std::make_unique<VideoCommon::GPUSynch>(system, use_nvdec);
-    }
    auto context = emu_window.CreateSharedContext();
    const auto scope = context->Acquire();
diff --git a/src/yuzu/applets/controller.cpp b/src/yuzu/applets/controller.cpp
index a15e8ca2a..c680fd2c2 100644
--- a/src/yuzu/applets/controller.cpp
+++ b/src/yuzu/applets/controller.cpp
@@ -535,7 +535,7 @@ void QtControllerSelectorDialog::UpdateControllerState(std::size_t player_index)
    // This emulates a delay between disconnecting and reconnecting controllers as some games
    // do not respond to a change in controller type if it was instantaneous.
    using namespace std::chrono_literals;
-    std::this_thread::sleep_for(20ms);
+    std::this_thread::sleep_for(60ms);
    UpdateController(controller_type, player_index, player_connected);
 }
diff --git a/src/yuzu/applets/error.cpp b/src/yuzu/applets/error.cpp
index 53a993cf6..8ee03ddb3 100644
--- a/src/yuzu/applets/error.cpp
+++ b/src/yuzu/applets/error.cpp
@@ -19,7 +19,7 @@ QtErrorDisplay::~QtErrorDisplay() = default;
 void QtErrorDisplay::ShowError(ResultCode error, std::function<void()> finished) const {
    callback = std::move(finished);
    emit MainWindowDisplayError(
-        tr("An error has occured.\nPlease try again or contact the developer of the "
+        tr("An error has occurred.\nPlease try again or contact the developer of the "
           "software.\n\nError Code: %1-%2 (0x%3)")
            .arg(static_cast<u32>(error.module.Value()) + 2000, 4, 10, QChar::fromLatin1('0'))
            .arg(error.description, 4, 10, QChar::fromLatin1('0'))
@@ -32,7 +32,7 @@ void QtErrorDisplay::ShowErrorWithTimestamp(ResultCode error, std::chrono::secon
    const QDateTime date_time = QDateTime::fromSecsSinceEpoch(time.count());
    emit MainWindowDisplayError(
-        tr("An error occured on %1 at %2.\nPlease try again or contact the "
+        tr("An error occurred on %1 at %2.\nPlease try again or contact the "
           "developer of the software.\n\nError Code: %3-%4 (0x%5)")
            .arg(date_time.toString(QStringLiteral("dddd, MMMM d, yyyy")))
            .arg(date_time.toString(QStringLiteral("h:mm:ss A")))
@@ -46,7 +46,7 @@ void QtErrorDisplay::ShowCustomErrorText(ResultCode error, std::string dialog_te
                                         std::function<void()> finished) const {
    callback = std::move(finished);
    emit MainWindowDisplayError(
-        tr("An error has occured.\nError Code: %1-%2 (0x%3)\n\n%4\n\n%5")
+        tr("An error has occurred.\nError Code: %1-%2 (0x%3)\n\n%4\n\n%5")
            .arg(static_cast<u32>(error.module.Value()) + 2000, 4, 10, QChar::fromLatin1('0'))
            .arg(error.description, 4, 10, QChar::fromLatin1('0'))
            .arg(error.raw, 8, 16, QChar::fromLatin1('0'))
diff --git a/src/yuzu/compatdb.cpp b/src/yuzu/compatdb.cpp
index 649912557..a470056ef 100644
--- a/src/yuzu/compatdb.cpp
+++ b/src/yuzu/compatdb.cpp
@@ -72,7 +72,7 @@ void CompatDB::Submit() {
 void CompatDB::OnTestcaseSubmitted() {
    if (!testcase_watcher.result()) {
        QMessageBox::critical(this, tr("Communication error"),
-                              tr("An error occured while sending the Testcase"));
+                              tr("An error occurred while sending the Testcase"));
        button(NextButton)->setEnabled(true);
        button(NextButton)->setText(tr("Next"));
        button(CancelButton)->setVisible(true);
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index 9fb254986..43cd11ba0 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -764,6 +764,8 @@ void Config::ReadCpuValues() {
            ReadSetting(QStringLiteral("cpuopt_unsafe_unfuse_fma"), true).toBool();
        Settings::values.cpuopt_unsafe_reduce_fp_error =
            ReadSetting(QStringLiteral("cpuopt_unsafe_reduce_fp_error"), true).toBool();
+        Settings::values.cpuopt_unsafe_inaccurate_nan =
+            ReadSetting(QStringLiteral("cpuopt_unsafe_inaccurate_nan"), true).toBool();
    }
    qt_config->endGroup();
@@ -1327,6 +1329,8 @@ void Config::SaveCpuValues() {
                     Settings::values.cpuopt_unsafe_unfuse_fma, true);
        WriteSetting(QStringLiteral("cpuopt_unsafe_reduce_fp_error"),
                     Settings::values.cpuopt_unsafe_reduce_fp_error, true);
+        WriteSetting(QStringLiteral("cpuopt_unsafe_inaccurate_nan"),
+                     Settings::values.cpuopt_unsafe_inaccurate_nan, true);
    }
    qt_config->endGroup();
@@ -1589,14 +1593,12 @@ void Config::WriteSettingGlobal(const QString& name, const QVariant& value, bool
 void Config::Reload() {
    ReadValues();
-    Settings::Sanitize();
    // To apply default value changes
    SaveValues();
    Settings::Apply(Core::System::GetInstance());
 }
 void Config::Save() {
-    Settings::Sanitize();
    SaveValues();
 }
diff --git a/src/yuzu/configuration/configure_cpu.cpp b/src/yuzu/configuration/configure_cpu.cpp
index 37fcd6adc..d055cbd60 100644
--- a/src/yuzu/configuration/configure_cpu.cpp
+++ b/src/yuzu/configuration/configure_cpu.cpp
@@ -36,6 +36,8 @@ void ConfigureCpu::SetConfiguration() {
    ui->cpuopt_unsafe_unfuse_fma->setChecked(Settings::values.cpuopt_unsafe_unfuse_fma);
    ui->cpuopt_unsafe_reduce_fp_error->setEnabled(runtime_lock);
    ui->cpuopt_unsafe_reduce_fp_error->setChecked(Settings::values.cpuopt_unsafe_reduce_fp_error);
+    ui->cpuopt_unsafe_inaccurate_nan->setEnabled(runtime_lock);
+    ui->cpuopt_unsafe_inaccurate_nan->setChecked(Settings::values.cpuopt_unsafe_inaccurate_nan);
 }
 void ConfigureCpu::AccuracyUpdated(int index) {
@@ -61,6 +63,7 @@ void ConfigureCpu::ApplyConfiguration() {
        static_cast<Settings::CPUAccuracy>(ui->accuracy->currentIndex());
    Settings::values.cpuopt_unsafe_unfuse_fma = ui->cpuopt_unsafe_unfuse_fma->isChecked();
    Settings::values.cpuopt_unsafe_reduce_fp_error = ui->cpuopt_unsafe_reduce_fp_error->isChecked();
+    Settings::values.cpuopt_unsafe_inaccurate_nan = ui->cpuopt_unsafe_inaccurate_nan->isChecked();
 }
 void ConfigureCpu::changeEvent(QEvent* event) {
diff --git a/src/yuzu/configuration/configure_cpu.ui b/src/yuzu/configuration/configure_cpu.ui
index ebdd2e6e9..bcd0962e9 100644
--- a/src/yuzu/configuration/configure_cpu.ui
+++ b/src/yuzu/configuration/configure_cpu.ui
@@ -109,6 +109,18 @@
          </property>
         </widget>
        </item>
+        <item>
+         <widget class="QCheckBox" name="cpuopt_unsafe_inaccurate_nan">
+          <property name="text">
+           <string>Inaccurate NaN handling</string>
+          </property>
+          <property name="toolTip">
+           <string>
+            &lt;div&gt;This option improves speed by removing NaN checking. Please note this also reduces accuracy of certain floating-point instructions.&lt;/div&gt;
+           </string>
+          </property>
+         </widget>
+        </item>
       </layout>
      </widget>
     </item>
diff --git a/src/yuzu/configuration/configure_input.cpp b/src/yuzu/configuration/configure_input.cpp
index d9009091b..567a36d9b 100644
--- a/src/yuzu/configuration/configure_input.cpp
+++ b/src/yuzu/configuration/configure_input.cpp
@@ -4,6 +4,7 @@
 #include <algorithm>
 #include <memory>
+#include <thread>
 #include <QSignalBlocker>
 #include <QTimer>
@@ -181,8 +182,18 @@ QList<QWidget*> ConfigureInput::GetSubTabs() const {
 }
 void ConfigureInput::ApplyConfiguration() {
-    for (auto controller : player_controllers) {
+    for (auto* controller : player_controllers) {
        controller->ApplyConfiguration();
+        controller->TryDisconnectSelectedController();
+    }
+    // This emulates a delay between disconnecting and reconnecting controllers as some games
+    // do not respond to a change in controller type if it was instantaneous.
+    using namespace std::chrono_literals;
+    std::this_thread::sleep_for(60ms);
+    for (auto* controller : player_controllers) {
+        controller->TryConnectSelectedController();
    }
    advanced->ApplyConfiguration();
diff --git a/src/yuzu/configuration/configure_input_player.cpp b/src/yuzu/configuration/configure_input_player.cpp
index 3c7500ee3..46ea026e4 100644
--- a/src/yuzu/configuration/configure_input_player.cpp
+++ b/src/yuzu/configuration/configure_input_player.cpp
@@ -4,7 +4,6 @@
 #include <algorithm>
 #include <memory>
-#include <thread>
 #include <utility>
 #include <QGridLayout>
 #include <QInputDialog>
@@ -576,6 +575,10 @@ void ConfigureInputPlayer::ApplyConfiguration() {
    std::transform(motions_param.begin(), motions_param.end(), motions.begin(),
                   [](const Common::ParamPackage& param) { return param.Serialize(); });
+}
+void ConfigureInputPlayer::TryConnectSelectedController() {
+    auto& player = Settings::values.players.GetValue()[player_index];
    const auto controller_type =
        GetControllerTypeFromIndex(ui->comboControllerType->currentIndex());
@@ -588,15 +591,12 @@ void ConfigureInputPlayer::ApplyConfiguration() {
        return;
    }
-    // Disconnect the controller first.
-    UpdateController(controller_type, player_index, false);
    player.controller_type = controller_type;
    player.connected = player_connected;
    ConfigureVibration::SetVibrationDevices(player_index);
-    // Handheld
+    // Connect/Disconnect Handheld depending on Player 1's controller configuration.
    if (player_index == 0) {
        auto& handheld = Settings::values.players.GetValue()[HANDHELD_INDEX];
        if (controller_type == Settings::ControllerType::Handheld) {
@@ -611,14 +611,26 @@ void ConfigureInputPlayer::ApplyConfiguration() {
        return;
    }
-    // This emulates a delay between disconnecting and reconnecting controllers as some games
-    // do not respond to a change in controller type if it was instantaneous.
-    using namespace std::chrono_literals;
-    std::this_thread::sleep_for(20ms);
    UpdateController(controller_type, player_index, player_connected);
 }
+void ConfigureInputPlayer::TryDisconnectSelectedController() {
+    const auto& player = Settings::values.players.GetValue()[player_index];
+    const auto controller_type =
+        GetControllerTypeFromIndex(ui->comboControllerType->currentIndex());
+    const auto player_connected = ui->groupConnectedController->isChecked() &&
+                                  controller_type != Settings::ControllerType::Handheld;
+    // Do not do anything if the controller configuration has not changed.
+    if (player.controller_type == controller_type && player.connected == player_connected) {
+        return;
+    }
+    // Disconnect the controller first.
+    UpdateController(controller_type, player_index, false);
+}
 void ConfigureInputPlayer::showEvent(QShowEvent* event) {
    if (bottom_row == nullptr) {
        return;
diff --git a/src/yuzu/configuration/configure_input_player.h b/src/yuzu/configuration/configure_input_player.h
index 9c30879a2..c4ae50de7 100644
--- a/src/yuzu/configuration/configure_input_player.h
+++ b/src/yuzu/configuration/configure_input_player.h
@@ -54,6 +54,18 @@ public:
    /// Save all button configurations to settings file.
    void ApplyConfiguration();
+    /**
+     * Attempts to connect the currently selected controller in the HID backend.
+     * This function will not do anything if it is not connected in the frontend.
+     */
+    void TryConnectSelectedController();
+    /**
+     * Attempts to disconnect the currently selected controller in the HID backend.
+     * This function will not do anything if the configuration has not changed.
+     */
+    void TryDisconnectSelectedController();
    /// Set the connection state checkbox (used to sync state).
    void ConnectPlayer(bool connected);
diff --git a/src/yuzu/main.cpp b/src/yuzu/main.cpp
index 7aa515226..43d64b708 100644
--- a/src/yuzu/main.cpp
+++ b/src/yuzu/main.cpp
@@ -142,7 +142,7 @@ constexpr int default_mouse_timeout = 2500;
 /**
 * "Callouts" are one-time instructional messages shown to the user. In the config settings, there
 * is a bitfield "callout_flags" options, used to track if a message has already been shown to the
- * user. This is 32-bits - if we have more than 32 callouts, we should retire and recyle old ones.
+ * user. This is 32-bits - if we have more than 32 callouts, we should retire and recycle old ones.
 */
 enum class CalloutFlag : uint32_t {
    Telemetry = 0x1,
@@ -580,9 +580,8 @@ void GMainWindow::InitializeWidgets() {
        if (emulation_running) {
            return;
        }
-        const bool is_async = !Settings::values.use_asynchronous_gpu_emulation.GetValue() ||
+        Settings::values.use_asynchronous_gpu_emulation.SetValue(
-                              Settings::values.use_multi_core.GetValue();
+            !Settings::values.use_asynchronous_gpu_emulation.GetValue());
-        Settings::values.use_asynchronous_gpu_emulation.SetValue(is_async);
        async_status_button->setChecked(Settings::values.use_asynchronous_gpu_emulation.GetValue());
        Settings::Apply(Core::System::GetInstance());
    });
@@ -599,16 +598,13 @@ void GMainWindow::InitializeWidgets() {
            return;
        }
        Settings::values.use_multi_core.SetValue(!Settings::values.use_multi_core.GetValue());
-        const bool is_async = Settings::values.use_asynchronous_gpu_emulation.GetValue() ||
-                              Settings::values.use_multi_core.GetValue();
-        Settings::values.use_asynchronous_gpu_emulation.SetValue(is_async);
-        async_status_button->setChecked(Settings::values.use_asynchronous_gpu_emulation.GetValue());
        multicore_status_button->setChecked(Settings::values.use_multi_core.GetValue());
        Settings::Apply(Core::System::GetInstance());
    });
    multicore_status_button->setText(tr("MULTICORE"));
    multicore_status_button->setCheckable(true);
    multicore_status_button->setChecked(Settings::values.use_multi_core.GetValue());
    statusBar()->insertPermanentWidget(0, multicore_status_button);
    statusBar()->insertPermanentWidget(0, async_status_button);
@@ -1049,20 +1045,23 @@ bool GMainWindow::LoadROM(const QString& filename, std::size_t program_index) {
            break;
        default:
-            if (static_cast<u32>(result) >
+            if (result > Core::System::ResultStatus::ErrorLoader) {
-                static_cast<u32>(Core::System::ResultStatus::ErrorLoader)) {
                const u16 loader_id = static_cast<u16>(Core::System::ResultStatus::ErrorLoader);
                const u16 error_id = static_cast<u16>(result) - loader_id;
                const std::string error_code = fmt::format("({:04X}-{:04X})", loader_id, error_id);
                LOG_CRITICAL(Frontend, "Failed to load ROM! {}", error_code);
-                QMessageBox::critical(
-                    this,
+                const auto title =
-                    tr("Error while loading ROM! ").append(QString::fromStdString(error_code)),
+                    tr("Error while loading ROM! %1", "%1 signifies a numeric error code.")
-                    QString::fromStdString(fmt::format(
+                        .arg(QString::fromStdString(error_code));
-                        "{}<br>Please follow <a href='https://yuzu-emu.org/help/quickstart/'>the "
+                const auto description =
-                        "yuzu quickstart guide</a> to redump your files.<br>You can refer "
+                    tr("%1<br>Please follow <a href='https://yuzu-emu.org/help/quickstart/'>the "
-                        "to the yuzu wiki</a> or the yuzu Discord</a> for help.",
+                       "yuzu quickstart guide</a> to redump your files.<br>You can refer "
-                        static_cast<Loader::ResultStatus>(error_id))));
+                       "to the yuzu wiki</a> or the yuzu Discord</a> for help.",
+                       "%1 signifies a numeric error ID.")
+                        .arg(error_id);
+                QMessageBox::critical(this, title, description);
            } else {
                QMessageBox::critical(
                    this, tr("Error while loading ROM!"),
@@ -2533,9 +2532,6 @@ void GMainWindow::UpdateStatusBar() {
 void GMainWindow::UpdateStatusButtons() {
    dock_status_button->setChecked(Settings::values.use_docked_mode.GetValue());
    multicore_status_button->setChecked(Settings::values.use_multi_core.GetValue());
-    Settings::values.use_asynchronous_gpu_emulation.SetValue(
-        Settings::values.use_asynchronous_gpu_emulation.GetValue() ||
-        Settings::values.use_multi_core.GetValue());
    async_status_button->setChecked(Settings::values.use_asynchronous_gpu_emulation.GetValue());
    renderer_status_button->setChecked(Settings::values.renderer_backend.GetValue() ==
                                       Settings::RendererBackend::Vulkan);
diff --git a/src/yuzu_cmd/yuzu.cpp b/src/yuzu_cmd/yuzu.cpp
index 2497c71ae..39e0d35aa 100644
--- a/src/yuzu_cmd/yuzu.cpp
+++ b/src/yuzu_cmd/yuzu.cpp
@@ -202,7 +202,7 @@ int main(int argc, char** argv) {
            const u16 loader_id = static_cast<u16>(Core::System::ResultStatus::ErrorLoader);
            const u16 error_id = static_cast<u16>(load_result) - loader_id;
            LOG_CRITICAL(Frontend,
-                         "While attempting to load the ROM requested, an error occured. Please "
+                         "While attempting to load the ROM requested, an error occurred. Please "
                         "refer to the yuzu wiki for more information or the yuzu discord for "
                         "additional help.\n\nError Code: {:04X}-{:04X}\nError Description: {}",
                         loader_id, error_id, static_cast<Loader::ResultStatus>(error_id));
diff --git a/src/yuzu_tester/yuzu.cpp b/src/yuzu_tester/yuzu.cpp
index 6435ffabb..09cf2ad77 100644
--- a/src/yuzu_tester/yuzu.cpp
+++ b/src/yuzu_tester/yuzu.cpp
@@ -242,7 +242,7 @@ int main(int argc, char** argv) {
            const u16 loader_id = static_cast<u16>(Core::System::ResultStatus::ErrorLoader);
            const u16 error_id = static_cast<u16>(load_result) - loader_id;
            LOG_CRITICAL(Frontend,
-                         "While attempting to load the ROM requested, an error occured. Please "
+                         "While attempting to load the ROM requested, an error occurred. Please "
                         "refer to the yuzu wiki for more information or the yuzu discord for "
                         "additional help.\n\nError Code: {:04X}-{:04X}\nError Description: {}",
                         loader_id, error_id, static_cast<Loader::ResultStatus>(error_id));