117 files changed, 2530 insertions, 337 deletions

diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt
index 0913be72c..3a57356ab 100644
--- a/src/CMakeLists.txt
+++ b/src/CMakeLists.txt
@@ -54,8 +54,10 @@ else()
     add_compile_options(
         -Wall
         -Werror=implicit-fallthrough
+        -Werror=missing-declarations
         -Werror=reorder
         -Wextra
+        -Wmissing-declarations
         -Wno-attributes
         -Wno-unused-parameter
     )
diff --git a/src/common/bit_field.h b/src/common/bit_field.h
index fd2bbbd99..26ae6c7fc 100644
--- a/src/common/bit_field.h
+++ b/src/common/bit_field.h
@@ -180,7 +180,7 @@ public:
     }
 
     constexpr void Assign(const T& value) {
-        storage = (static_cast<StorageType>(storage) & ~mask) | FormatValue(value);
+        storage = static_cast<StorageType>((storage & ~mask) | FormatValue(value));
     }
 
     constexpr T Value() const {
diff --git a/src/core/crypto/partition_data_manager.cpp b/src/core/crypto/partition_data_manager.cpp
index d64302f2e..7ed71ac3a 100644
--- a/src/core/crypto/partition_data_manager.cpp
+++ b/src/core/crypto/partition_data_manager.cpp
@@ -202,8 +202,8 @@ static std::array<Key128, 0x20> FindEncryptedMasterKeyFromHex(const std::vector<
     return out;
 }
 
-FileSys::VirtualFile FindFileInDirWithNames(const FileSys::VirtualDir& dir,
-                                            const std::string& name) {
+static FileSys::VirtualFile FindFileInDirWithNames(const FileSys::VirtualDir& dir,
+                                                   const std::string& name) {
     const auto upper = Common::ToUpper(name);
 
     for (const auto& fname : {name, name + ".bin", upper, upper + ".BIN"}) {
@@ -345,8 +345,7 @@ FileSys::VirtualFile PartitionDataManager::GetPackage2Raw(Package2Type type) con
     return package2.at(static_cast<size_t>(type));
 }
 
-bool AttemptDecrypt(const std::array<u8, 16>& key, Package2Header& header) {
-
+static bool AttemptDecrypt(const std::array<u8, 16>& key, Package2Header& header) {
     const std::vector<u8> iv(header.header_ctr.begin(), header.header_ctr.end());
     Package2Header temp = header;
     AESCipher<Key128> cipher(key, Mode::CTR);
diff --git a/src/core/gdbstub/gdbstub.cpp b/src/core/gdbstub/gdbstub.cpp
index 2f15635c5..70c0f8b80 100644
--- a/src/core/gdbstub/gdbstub.cpp
+++ b/src/core/gdbstub/gdbstub.cpp
@@ -1389,10 +1389,9 @@ void SendTrap(Kernel::Thread* thread, int trap) {
         return;
     }
 
-    if (!halt_loop || current_thread == thread) {
-        current_thread = thread;
-        SendSignal(thread, trap);
-    }
+    current_thread = thread;
+    SendSignal(thread, trap);
+
     halt_loop = true;
     send_trap = false;
 }
diff --git a/src/core/hle/kernel/memory/memory_block.h b/src/core/hle/kernel/memory/memory_block.h
index e11043b60..9db1f7b39 100644
--- a/src/core/hle/kernel/memory/memory_block.h
+++ b/src/core/hle/kernel/memory/memory_block.h
@@ -17,7 +17,7 @@ namespace Kernel::Memory {
 
 enum class MemoryState : u32 {
     None = 0,
-    Mask = 0xFFFFFFFF, // TODO(bunnei): This should probable be 0xFF
+    Mask = 0xFF,
     All = ~None,
 
     FlagCanReprotect = (1 << 8),
@@ -253,6 +253,23 @@ public:
         };
     }
 
+    void ShareToDevice(MemoryPermission /*new_perm*/) {
+        ASSERT((attribute & MemoryAttribute::DeviceShared) == MemoryAttribute::DeviceShared ||
+               device_use_count == 0);
+        attribute |= MemoryAttribute::DeviceShared;
+        const u16 new_use_count{++device_use_count};
+        ASSERT(new_use_count > 0);
+    }
+
+    void UnshareToDevice(MemoryPermission /*new_perm*/) {
+        ASSERT((attribute & MemoryAttribute::DeviceShared) == MemoryAttribute::DeviceShared);
+        const u16 prev_use_count{device_use_count--};
+        ASSERT(prev_use_count > 0);
+        if (prev_use_count == 1) {
+            attribute &= ~MemoryAttribute::DeviceShared;
+        }
+    }
+
 private:
     constexpr bool HasProperties(MemoryState s, MemoryPermission p, MemoryAttribute a) const {
         constexpr MemoryAttribute AttributeIgnoreMask{MemoryAttribute::DontCareMask |
@@ -287,9 +304,9 @@ private:
         state = new_state;
         perm = new_perm;
 
-        // TODO(bunnei): Is this right?
         attribute = static_cast<MemoryAttribute>(
-            new_attribute /*| (attribute & (MemoryAttribute::IpcLocked | MemoryAttribute::DeviceShared))*/);
+            new_attribute |
+            (attribute & (MemoryAttribute::IpcLocked | MemoryAttribute::DeviceShared)));
     }
 
     constexpr MemoryBlock Split(VAddr split_addr) {
diff --git a/src/core/hle/kernel/memory/memory_block_manager.cpp b/src/core/hle/kernel/memory/memory_block_manager.cpp
index 1ebc126c0..900395c37 100644
--- a/src/core/hle/kernel/memory/memory_block_manager.cpp
+++ b/src/core/hle/kernel/memory/memory_block_manager.cpp
@@ -143,6 +143,42 @@ void MemoryBlockManager::Update(VAddr addr, std::size_t num_pages, MemoryState s
     }
 }
 
+void MemoryBlockManager::UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func,
+                                    MemoryPermission perm) {
+    const std::size_t prev_count{memory_block_tree.size()};
+    const VAddr end_addr{addr + num_pages * PageSize};
+    iterator node{memory_block_tree.begin()};
+
+    while (node != memory_block_tree.end()) {
+        MemoryBlock* block{&(*node)};
+        iterator next_node{std::next(node)};
+        const VAddr cur_addr{block->GetAddress()};
+        const VAddr cur_end_addr{block->GetNumPages() * PageSize + cur_addr};
+
+        if (addr < cur_end_addr && cur_addr < end_addr) {
+            iterator new_node{node};
+
+            if (addr > cur_addr) {
+                memory_block_tree.insert(node, block->Split(addr));
+            }
+
+            if (end_addr < cur_end_addr) {
+                new_node = memory_block_tree.insert(node, block->Split(end_addr));
+            }
+
+            lock_func(new_node, perm);
+
+            MergeAdjacent(new_node, next_node);
+        }
+
+        if (cur_end_addr - 1 >= end_addr - 1) {
+            break;
+        }
+
+        node = next_node;
+    }
+}
+
 void MemoryBlockManager::IterateForRange(VAddr start, VAddr end, IterateFunc&& func) {
     const_iterator it{FindIterator(start)};
     MemoryInfo info{};
diff --git a/src/core/hle/kernel/memory/memory_block_manager.h b/src/core/hle/kernel/memory/memory_block_manager.h
index 0f2270f0f..9451b5df6 100644
--- a/src/core/hle/kernel/memory/memory_block_manager.h
+++ b/src/core/hle/kernel/memory/memory_block_manager.h
@@ -45,6 +45,9 @@ public:
                 MemoryPermission perm = MemoryPermission::None,
                 MemoryAttribute attribute = MemoryAttribute::None);
 
+    using LockFunc = std::function<void(iterator, MemoryPermission)>;
+    void UpdateLock(VAddr addr, std::size_t num_pages, LockFunc&& lock_func, MemoryPermission perm);
+
     using IterateFunc = std::function<void(const MemoryInfo&)>;
     void IterateForRange(VAddr start, VAddr end, IterateFunc&& func);
 
diff --git a/src/core/hle/kernel/memory/page_table.cpp b/src/core/hle/kernel/memory/page_table.cpp
index 091e52ca4..3281611f8 100644
--- a/src/core/hle/kernel/memory/page_table.cpp
+++ b/src/core/hle/kernel/memory/page_table.cpp
@@ -840,6 +840,50 @@ ResultVal<VAddr> PageTable::AllocateAndMapMemory(std::size_t needed_num_pages, s
     return MakeResult<VAddr>(addr);
 }
 
+ResultCode PageTable::LockForDeviceAddressSpace(VAddr addr, std::size_t size) {
+    std::lock_guard lock{page_table_lock};
+
+    MemoryPermission perm{};
+    if (const ResultCode result{CheckMemoryState(
+            nullptr, &perm, nullptr, addr, size, MemoryState::FlagCanChangeAttribute,
+            MemoryState::FlagCanChangeAttribute, MemoryPermission::None, MemoryPermission::None,
+            MemoryAttribute::LockedAndIpcLocked, MemoryAttribute::None,
+            MemoryAttribute::DeviceSharedAndUncached)};
+        result.IsError()) {
+        return result;
+    }
+
+    block_manager->UpdateLock(addr, size / PageSize,
+                              [](MemoryBlockManager::iterator block, MemoryPermission perm) {
+                                  block->ShareToDevice(perm);
+                              },
+                              perm);
+
+    return RESULT_SUCCESS;
+}
+
+ResultCode PageTable::UnlockForDeviceAddressSpace(VAddr addr, std::size_t size) {
+    std::lock_guard lock{page_table_lock};
+
+    MemoryPermission perm{};
+    if (const ResultCode result{CheckMemoryState(
+            nullptr, &perm, nullptr, addr, size, MemoryState::FlagCanChangeAttribute,
+            MemoryState::FlagCanChangeAttribute, MemoryPermission::None, MemoryPermission::None,
+            MemoryAttribute::LockedAndIpcLocked, MemoryAttribute::None,
+            MemoryAttribute::DeviceSharedAndUncached)};
+        result.IsError()) {
+        return result;
+    }
+
+    block_manager->UpdateLock(addr, size / PageSize,
+                              [](MemoryBlockManager::iterator block, MemoryPermission perm) {
+                                  block->UnshareToDevice(perm);
+                              },
+                              perm);
+
+    return RESULT_SUCCESS;
+}
+
 ResultCode PageTable::InitializeMemoryLayout(VAddr start, VAddr end) {
     block_manager = std::make_unique<MemoryBlockManager>(start, end);
 
diff --git a/src/core/hle/kernel/memory/page_table.h b/src/core/hle/kernel/memory/page_table.h
index 80384ab0f..a867aa050 100644
--- a/src/core/hle/kernel/memory/page_table.h
+++ b/src/core/hle/kernel/memory/page_table.h
@@ -53,6 +53,8 @@ public:
                                           bool is_map_only, VAddr region_start,
                                           std::size_t region_num_pages, MemoryState state,
                                           MemoryPermission perm, PAddr map_addr = 0);
+    ResultCode LockForDeviceAddressSpace(VAddr addr, std::size_t size);
+    ResultCode UnlockForDeviceAddressSpace(VAddr addr, std::size_t size);
 
     Common::PageTable& PageTableImpl() {
         return page_table_impl;
diff --git a/src/core/hle/kernel/shared_memory.cpp b/src/core/hle/kernel/shared_memory.cpp
index c67696757..0cd467110 100644
--- a/src/core/hle/kernel/shared_memory.cpp
+++ b/src/core/hle/kernel/shared_memory.cpp
@@ -36,22 +36,22 @@ std::shared_ptr<SharedMemory> SharedMemory::Create(
 }
 
 ResultCode SharedMemory::Map(Process& target_process, VAddr address, std::size_t size,
-                             Memory::MemoryPermission permission) {
+                             Memory::MemoryPermission permissions) {
     const u64 page_count{(size + Memory::PageSize - 1) / Memory::PageSize};
 
     if (page_list.GetNumPages() != page_count) {
         UNIMPLEMENTED_MSG("Page count does not match");
     }
 
-    Memory::MemoryPermission expected =
+    const Memory::MemoryPermission expected =
         &target_process == owner_process ? owner_permission : user_permission;
 
-    if (permission != expected) {
+    if (permissions != expected) {
         UNIMPLEMENTED_MSG("Permission does not match");
     }
 
     return target_process.PageTable().MapPages(address, page_list, Memory::MemoryState::Shared,
-                                               permission);
+                                               permissions);
 }
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/shared_memory.h b/src/core/hle/kernel/shared_memory.h
index cd16d6412..0ef87235c 100644
--- a/src/core/hle/kernel/shared_memory.h
+++ b/src/core/hle/kernel/shared_memory.h
@@ -51,7 +51,7 @@ public:
      * @param permissions Memory block map permissions (specified by SVC field)
      */
     ResultCode Map(Process& target_process, VAddr address, std::size_t size,
-                   Memory::MemoryPermission permission);
+                   Memory::MemoryPermission permissions);
 
     /**
      * Gets a pointer to the shared memory block
diff --git a/src/core/hle/kernel/svc.cpp b/src/core/hle/kernel/svc.cpp
index 4134acf65..25b4a23b4 100644
--- a/src/core/hle/kernel/svc.cpp
+++ b/src/core/hle/kernel/svc.cpp
@@ -55,9 +55,6 @@ constexpr bool IsValidAddressRange(VAddr address, u64 size) {
     return address + size > address;
 }
 
-// 8 GiB
-constexpr u64 MAIN_MEMORY_SIZE = 0x200000000;
-
 // Helper function that performs the common sanity checks for svcMapMemory
 // and svcUnmapMemory. This is doable, as both functions perform their sanitizing
 // in the same order.
@@ -1229,6 +1226,142 @@ static ResultCode QueryMemory32(Core::System& system, u32 memory_info_address,
     return QueryMemory(system, memory_info_address, page_info_address, query_address);
 }
 
+static ResultCode MapProcessCodeMemory(Core::System& system, Handle process_handle, u64 dst_address,
+                                       u64 src_address, u64 size) {
+    LOG_DEBUG(Kernel_SVC,
+              "called. process_handle=0x{:08X}, dst_address=0x{:016X}, "
+              "src_address=0x{:016X}, size=0x{:016X}",
+              process_handle, dst_address, src_address, size);
+
+    if (!Common::Is4KBAligned(src_address)) {
+        LOG_ERROR(Kernel_SVC, "src_address is not page-aligned (src_address=0x{:016X}).",
+                  src_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (!Common::Is4KBAligned(dst_address)) {
+        LOG_ERROR(Kernel_SVC, "dst_address is not page-aligned (dst_address=0x{:016X}).",
+                  dst_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (size == 0 || !Common::Is4KBAligned(size)) {
+        LOG_ERROR(Kernel_SVC, "Size is zero or not page-aligned (size=0x{:016X})", size);
+        return ERR_INVALID_SIZE;
+    }
+
+    if (!IsValidAddressRange(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range overflows the address space (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!IsValidAddressRange(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range overflows the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
+    auto process = handle_table.Get<Process>(process_handle);
+    if (!process) {
+        LOG_ERROR(Kernel_SVC, "Invalid process handle specified (handle=0x{:08X}).",
+                  process_handle);
+        return ERR_INVALID_HANDLE;
+    }
+
+    auto& page_table = process->PageTable();
+    if (!page_table.IsInsideAddressSpace(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range is not within the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!page_table.IsInsideASLRRegion(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range is not within the ASLR region (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_MEMORY_RANGE;
+    }
+
+    return page_table.MapProcessCodeMemory(dst_address, src_address, size);
+}
+
+static ResultCode UnmapProcessCodeMemory(Core::System& system, Handle process_handle,
+                                         u64 dst_address, u64 src_address, u64 size) {
+    LOG_DEBUG(Kernel_SVC,
+              "called. process_handle=0x{:08X}, dst_address=0x{:016X}, src_address=0x{:016X}, "
+              "size=0x{:016X}",
+              process_handle, dst_address, src_address, size);
+
+    if (!Common::Is4KBAligned(dst_address)) {
+        LOG_ERROR(Kernel_SVC, "dst_address is not page-aligned (dst_address=0x{:016X}).",
+                  dst_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (!Common::Is4KBAligned(src_address)) {
+        LOG_ERROR(Kernel_SVC, "src_address is not page-aligned (src_address=0x{:016X}).",
+                  src_address);
+        return ERR_INVALID_ADDRESS;
+    }
+
+    if (size == 0 || Common::Is4KBAligned(size)) {
+        LOG_ERROR(Kernel_SVC, "Size is zero or not page-aligned (size=0x{:016X}).", size);
+        return ERR_INVALID_SIZE;
+    }
+
+    if (!IsValidAddressRange(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range overflows the address space (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!IsValidAddressRange(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range overflows the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    const auto& handle_table = system.Kernel().CurrentProcess()->GetHandleTable();
+    auto process = handle_table.Get<Process>(process_handle);
+    if (!process) {
+        LOG_ERROR(Kernel_SVC, "Invalid process handle specified (handle=0x{:08X}).",
+                  process_handle);
+        return ERR_INVALID_HANDLE;
+    }
+
+    auto& page_table = process->PageTable();
+    if (!page_table.IsInsideAddressSpace(src_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Source address range is not within the address space (src_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  src_address, size);
+        return ERR_INVALID_ADDRESS_STATE;
+    }
+
+    if (!page_table.IsInsideASLRRegion(dst_address, size)) {
+        LOG_ERROR(Kernel_SVC,
+                  "Destination address range is not within the ASLR region (dst_address=0x{:016X}, "
+                  "size=0x{:016X}).",
+                  dst_address, size);
+        return ERR_INVALID_MEMORY_RANGE;
+    }
+
+    return page_table.UnmapProcessCodeMemory(dst_address, src_address, size);
+}
+
 /// Exits the current process
 static void ExitProcess(Core::System& system) {
     auto* current_process = system.Kernel().CurrentProcess();
@@ -2256,8 +2389,8 @@ static const FunctionDef SVC_Table_64[] = {
     {0x74, nullptr, "MapProcessMemory"},
     {0x75, nullptr, "UnmapProcessMemory"},
     {0x76, SvcWrap64<QueryProcessMemory>, "QueryProcessMemory"},
-    {0x77, nullptr, "MapProcessCodeMemory"},
-    {0x78, nullptr, "UnmapProcessCodeMemory"},
+    {0x77, SvcWrap64<MapProcessCodeMemory>, "MapProcessCodeMemory"},
+    {0x78, SvcWrap64<UnmapProcessCodeMemory>, "UnmapProcessCodeMemory"},
     {0x79, nullptr, "CreateProcess"},
     {0x7A, nullptr, "StartProcess"},
     {0x7B, nullptr, "TerminateProcess"},
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index 4c0451c01..a919750a6 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -150,8 +150,7 @@ static void ResetThreadContext64(Core::ARM_Interface::ThreadContext64& context,
     context.pc = entry_point;
     context.sp = stack_top;
     // TODO(merry): Perform a hardware test to determine the below value.
-    // AHP = 0, DN = 1, FTZ = 1, RMode = Round towards zero
-    context.fpcr = 0x03C00000;
+    context.fpcr = 0;
 }
 
 ResultVal<std::shared_ptr<Thread>> Thread::Create(KernelCore& kernel, std::string name,
diff --git a/src/core/hle/service/acc/acc_su.cpp b/src/core/hle/service/acc/acc_su.cpp
index b941c260b..ae88deda5 100644
--- a/src/core/hle/service/acc/acc_su.cpp
+++ b/src/core/hle/service/acc/acc_su.cpp
@@ -33,8 +33,10 @@ ACC_SU::ACC_SU(std::shared_ptr<Module> module, std::shared_ptr<ProfileManager> p
         {111, nullptr, "ClearSaveDataThumbnail"},
         {112, nullptr, "LoadSaveDataThumbnail"},
         {113, nullptr, "GetSaveDataThumbnailExistence"},
+        {120, nullptr, "ListOpenUsersInApplication"},
         {130, nullptr, "ActivateOpenContextRetention"},
         {140, nullptr, "ListQualifiedUsers"},
+        {150, nullptr, "AuthenticateApplicationAsync"},
         {190, nullptr, "GetUserLastOpenedApplication"},
         {191, nullptr, "ActivateOpenContextHolder"},
         {200, nullptr, "BeginUserRegistration"},
diff --git a/src/core/hle/service/acc/acc_u1.cpp b/src/core/hle/service/acc/acc_u1.cpp
index 858e91dde..2b9c11928 100644
--- a/src/core/hle/service/acc/acc_u1.cpp
+++ b/src/core/hle/service/acc/acc_u1.cpp
@@ -35,6 +35,7 @@ ACC_U1::ACC_U1(std::shared_ptr<Module> module, std::shared_ptr<ProfileManager> p
         {113, nullptr, "GetSaveDataThumbnailExistence"},
         {130, nullptr, "ActivateOpenContextRetention"},
         {140, nullptr, "ListQualifiedUsers"},
+        {150, nullptr, "AuthenticateApplicationAsync"},
         {190, nullptr, "GetUserLastOpenedApplication"},
         {191, nullptr, "ActivateOpenContextHolder"},
         {997, nullptr, "DebugInvalidateTokenCacheForUser"},
diff --git a/src/core/hle/service/am/am.cpp b/src/core/hle/service/am/am.cpp
index 3ece2cf3c..bee4a9d3f 100644
--- a/src/core/hle/service/am/am.cpp
+++ b/src/core/hle/service/am/am.cpp
@@ -235,6 +235,7 @@ IDebugFunctions::IDebugFunctions() : ServiceFramework{"IDebugFunctions"} {
         {30, nullptr, "RequestLaunchApplicationWithUserAndArgumentForDebug"},
         {40, nullptr, "GetAppletResourceUsageInfo"},
         {100, nullptr, "SetCpuBoostModeForApplet"},
+        {101, nullptr, "CancelCpuBoostModeForApplet"},
         {110, nullptr, "PushToAppletBoundChannelForDebug"},
         {111, nullptr, "TryPopFromAppletBoundChannelForDebug"},
         {120, nullptr, "AlarmSettingNotificationEnableAppEventReserve"},
@@ -277,6 +278,8 @@ ISelfController::ISelfController(Core::System& system,
         {41, nullptr, "IsSystemBufferSharingEnabled"},
         {42, nullptr, "GetSystemSharedLayerHandle"},
         {43, nullptr, "GetSystemSharedBufferHandle"},
+        {44, nullptr, "CreateManagedDisplaySeparableLayer"},
+        {45, nullptr, "SetManagedDisplayLayerSeparationMode"},
         {50, &ISelfController::SetHandlesRequestToDisplay, "SetHandlesRequestToDisplay"},
         {51, nullptr, "ApproveToDisplay"},
         {60, nullptr, "OverrideAutoSleepTimeAndDimmingTime"},
@@ -623,11 +626,15 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system,
         {64, nullptr, "SetTvPowerStateMatchingMode"},
         {65, nullptr, "GetApplicationIdByContentActionName"},
         {66, &ICommonStateGetter::SetCpuBoostMode, "SetCpuBoostMode"},
+        {67, nullptr, "CancelCpuBoostMode"},
         {80, nullptr, "PerformSystemButtonPressingIfInFocus"},
         {90, nullptr, "SetPerformanceConfigurationChangedNotification"},
         {91, nullptr, "GetCurrentPerformanceConfiguration"},
+        {100, nullptr, "SetHandlingHomeButtonShortPressedEnabled"},
         {200, nullptr, "GetOperationModeSystemInfo"},
         {300, nullptr, "GetSettingsPlatformRegion"},
+        {400, nullptr, "ActivateMigrationService"},
+        {401, nullptr, "DeactivateMigrationService"},
     };
     // clang-format on
 
@@ -835,6 +842,7 @@ public:
             {25, nullptr, "Terminate"},
             {30, &ILibraryAppletAccessor::GetResult, "GetResult"},
             {50, nullptr, "SetOutOfFocusApplicationSuspendingEnabled"},
+            {60, nullptr, "PresetLibraryAppletGpuTimeSliceZero"},
             {100, &ILibraryAppletAccessor::PushInData, "PushInData"},
             {101, &ILibraryAppletAccessor::PopOutData, "PopOutData"},
             {102, nullptr, "PushExtraStorage"},
@@ -1139,6 +1147,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {31, &IApplicationFunctions::EndBlockingHomeButtonShortAndLongPressed, "EndBlockingHomeButtonShortAndLongPressed"},
         {32, &IApplicationFunctions::BeginBlockingHomeButton, "BeginBlockingHomeButton"},
         {33, &IApplicationFunctions::EndBlockingHomeButton, "EndBlockingHomeButton"},
+        {34, nullptr, "SelectApplicationLicense"},
         {40, &IApplicationFunctions::NotifyRunning, "NotifyRunning"},
         {50, &IApplicationFunctions::GetPseudoDeviceId, "GetPseudoDeviceId"},
         {60, nullptr, "SetMediaPlaybackStateForApplication"},
@@ -1148,6 +1157,7 @@ IApplicationFunctions::IApplicationFunctions(Core::System& system_)
         {68, nullptr, "RequestFlushGamePlayingMovieForDebug"},
         {70, nullptr, "RequestToShutdown"},
         {71, nullptr, "RequestToReboot"},
+        {72, nullptr, "RequestToSleep"},
         {80, nullptr, "ExitAndRequestToShowThanksMessage"},
         {90, &IApplicationFunctions::EnableApplicationCrashReport, "EnableApplicationCrashReport"},
         {100, &IApplicationFunctions::InitializeApplicationCopyrightFrameBuffer, "InitializeApplicationCopyrightFrameBuffer"},
diff --git a/src/core/hle/service/audio/audctl.cpp b/src/core/hle/service/audio/audctl.cpp
index 9e08e5346..6ddb547fb 100644
--- a/src/core/hle/service/audio/audctl.cpp
+++ b/src/core/hle/service/audio/audctl.cpp
@@ -39,6 +39,8 @@ AudCtl::AudCtl() : ServiceFramework{"audctl"} {
         {25, nullptr, "GetAudioVolumeDataForPlayReport"},
         {26, nullptr, "UpdateHeadphoneSettings"},
         {27, nullptr, "SetVolumeMappingTableForDev"},
+        {28, nullptr, "GetAudioOutputChannelCountForPlayReport"},
+        {29, nullptr, "BindAudioOutputChannelCountUpdateEventForPlayReport"},
     };
     // clang-format on
 
diff --git a/src/core/hle/service/bcat/backend/boxcat.cpp b/src/core/hle/service/bcat/backend/boxcat.cpp
index f589864ee..5febe8fc1 100644
--- a/src/core/hle/service/bcat/backend/boxcat.cpp
+++ b/src/core/hle/service/bcat/backend/boxcat.cpp
@@ -18,6 +18,7 @@
 #include "core/hle/service/bcat/backend/boxcat.h"
 #include "core/settings.h"
 
+namespace Service::BCAT {
 namespace {
 
 // Prevents conflicts with windows macro called CreateFile
@@ -30,10 +31,6 @@ bool VfsDeleteFileWrap(FileSys::VirtualDir dir, std::string_view name) {
     return dir->DeleteFile(name);
 }
 
-} // Anonymous namespace
-
-namespace Service::BCAT {
-
 constexpr ResultCode ERROR_GENERAL_BCAT_FAILURE{ErrorModule::BCAT, 1};
 
 constexpr char BOXCAT_HOSTNAME[] = "api.yuzu-emu.org";
@@ -90,8 +87,6 @@ constexpr u32 PORT = 443;
 constexpr u32 TIMEOUT_SECONDS = 30;
 [[maybe_unused]] constexpr u64 VFS_COPY_BLOCK_SIZE = 1ULL << 24; // 4MB
 
-namespace {
-
 std::string GetBINFilePath(u64 title_id) {
     return fmt::format("{}bcat/{:016X}/launchparam.bin",
                        FileUtil::GetUserPath(FileUtil::UserPath::CacheDir), title_id);
diff --git a/src/core/hle/service/bcat/module.cpp b/src/core/hle/service/bcat/module.cpp
index 7ada67130..34aba7a27 100644
--- a/src/core/hle/service/bcat/module.cpp
+++ b/src/core/hle/service/bcat/module.cpp
@@ -141,6 +141,7 @@ public:
             {20301, nullptr, "RequestSuspendDeliveryTask"},
             {20400, nullptr, "RegisterSystemApplicationDeliveryTask"},
             {20401, nullptr, "UnregisterSystemApplicationDeliveryTask"},
+            {20410, nullptr, "SetSystemApplicationDeliveryTaskTimer"},
             {30100, &IBcatService::SetPassphrase, "SetPassphrase"},
             {30200, nullptr, "RegisterBackgroundDeliveryTask"},
             {30201, nullptr, "UnregisterBackgroundDeliveryTask"},
diff --git a/src/core/hle/service/es/es.cpp b/src/core/hle/service/es/es.cpp
index 86f36915a..f8e9df4b1 100644
--- a/src/core/hle/service/es/es.cpp
+++ b/src/core/hle/service/es/es.cpp
@@ -4,6 +4,7 @@
 
 #include "core/crypto/key_manager.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/service/es/es.h"
 #include "core/hle/service/service.h"
 
 namespace Service::ES {
diff --git a/src/core/hle/service/filesystem/fsp_srv.cpp b/src/core/hle/service/filesystem/fsp_srv.cpp
index 61045c75c..6b9b4f3b9 100644
--- a/src/core/hle/service/filesystem/fsp_srv.cpp
+++ b/src/core/hle/service/filesystem/fsp_srv.cpp
@@ -697,12 +697,14 @@ FSP_SRV::FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter)
         {68, nullptr, "OpenSaveDataInfoReaderBySaveDataFilter"},
         {69, nullptr, "ReadSaveDataFileSystemExtraDataBySaveDataAttribute"},
         {70, nullptr, "WriteSaveDataFileSystemExtraDataBySaveDataAttribute"},
+        {71, nullptr, "ReadSaveDataFileSystemExtraDataWithMaskBySaveDataAttribute"},
         {80, nullptr, "OpenSaveDataMetaFile"},
         {81, nullptr, "OpenSaveDataTransferManager"},
         {82, nullptr, "OpenSaveDataTransferManagerVersion2"},
         {83, nullptr, "OpenSaveDataTransferProhibiterForCloudBackUp"},
         {84, nullptr, "ListApplicationAccessibleSaveDataOwnerId"},
         {85, nullptr, "OpenSaveDataTransferManagerForSaveDataRepair"},
+        {86, nullptr, "OpenSaveDataMover"},
         {100, nullptr, "OpenImageDirectoryFileSystem"},
         {110, nullptr, "OpenContentStorageFileSystem"},
         {120, nullptr, "OpenCloudBackupWorkStorageFileSystem"},
@@ -762,9 +764,11 @@ FSP_SRV::FSP_SRV(FileSystemController& fsc, const Core::Reporter& reporter)
         {1011, &FSP_SRV::GetAccessLogVersionInfo, "GetAccessLogVersionInfo"},
         {1012, nullptr, "GetFsStackUsage"},
         {1013, nullptr, "UnsetSaveDataRootPath"},
+        {1014, nullptr, "OutputMultiProgramTagAccessLog"},
         {1100, nullptr, "OverrideSaveDataTransferTokenSignVerificationKey"},
         {1110, nullptr, "CorruptSaveDataFileSystemBySaveDataSpaceId2"},
         {1200, nullptr, "OpenMultiCommitManager"},
+        {1300, nullptr, "OpenBisWiper"},
     };
     // clang-format on
     RegisterHandlers(functions);
diff --git a/src/core/hle/service/friend/friend.cpp b/src/core/hle/service/friend/friend.cpp
index 7938b4b80..68f259b70 100644
--- a/src/core/hle/service/friend/friend.cpp
+++ b/src/core/hle/service/friend/friend.cpp
@@ -96,6 +96,7 @@ public:
             {30830, nullptr, "ClearPlayLog"},
             {30900, nullptr, "SendFriendInvitation"},
             {30910, nullptr, "ReadFriendInvitation"},
+            {30911, nullptr, "ReadAllFriendInvitations"},
             {49900, nullptr, "DeleteNetworkServiceAccountCache"},
         };
         // clang-format on
diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp
index d6031a987..5559587e3 100644
--- a/src/core/hle/service/hid/hid.cpp
+++ b/src/core/hle/service/hid/hid.cpp
@@ -233,7 +233,7 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
         {302, nullptr, "StopConsoleSixAxisSensor"},
         {303, nullptr, "ActivateSevenSixAxisSensor"},
         {304, nullptr, "StartSevenSixAxisSensor"},
-        {305, nullptr, "StopSevenSixAxisSensor"},
+        {305, &Hid::StopSevenSixAxisSensor, "StopSevenSixAxisSensor"},
         {306, &Hid::InitializeSevenSixAxisSensor, "InitializeSevenSixAxisSensor"},
         {307, nullptr, "FinalizeSevenSixAxisSensor"},
         {308, nullptr, "SetSevenSixAxisSensorFusionStrength"},
@@ -282,6 +282,7 @@ Hid::Hid(Core::System& system) : ServiceFramework("hid"), system(system) {
         {1001, nullptr, "GetNpadCommunicationMode"},
         {1002, nullptr, "SetTouchScreenConfiguration"},
         {1003, nullptr, "IsFirmwareUpdateNeededForNotification"},
+        {2000, nullptr, "ActivateDigitizer"},
     };
     // clang-format on
 
@@ -852,6 +853,17 @@ void Hid::SetPalmaBoostMode(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
+void Hid::StopSevenSixAxisSensor(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const auto applet_resource_user_id{rp.Pop<u64>()};
+
+    LOG_WARNING(Service_HID, "(STUBBED) called, applet_resource_user_id={}",
+                applet_resource_user_id);
+
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+}
+
 void Hid::InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service_HID, "(STUBBED) called");
 
@@ -870,6 +882,7 @@ public:
             {10, nullptr, "DeactivateTouchScreen"},
             {11, nullptr, "SetTouchScreenAutoPilotState"},
             {12, nullptr, "UnsetTouchScreenAutoPilotState"},
+            {13, nullptr, "GetTouchScreenConfiguration"},
             {20, nullptr, "DeactivateMouse"},
             {21, nullptr, "SetMouseAutoPilotState"},
             {22, nullptr, "UnsetMouseAutoPilotState"},
@@ -879,7 +892,9 @@ public:
             {50, nullptr, "DeactivateXpad"},
             {51, nullptr, "SetXpadAutoPilotState"},
             {52, nullptr, "UnsetXpadAutoPilotState"},
-            {60, nullptr, "DeactivateJoyXpad"},
+            {60, nullptr, "ClearNpadSystemCommonPolicy"},
+            {61, nullptr, "DeactivateNpad"},
+            {62, nullptr, "ForceDisconnectNpad"},
             {91, nullptr, "DeactivateGesture"},
             {110, nullptr, "DeactivateHomeButton"},
             {111, nullptr, "SetHomeButtonAutoPilotState"},
@@ -899,6 +914,15 @@ public:
             {141, nullptr, "GetConsoleSixAxisSensorSamplingFrequency"},
             {142, nullptr, "DeactivateSevenSixAxisSensor"},
             {143, nullptr, "GetConsoleSixAxisSensorCountStates"},
+            {144, nullptr, "GetAccelerometerFsr"},
+            {145, nullptr, "SetAccelerometerFsr"},
+            {146, nullptr, "GetAccelerometerOdr"},
+            {147, nullptr, "SetAccelerometerOdr"},
+            {148, nullptr, "GetGyroscopeFsr"},
+            {149, nullptr, "SetGyroscopeFsr"},
+            {150, nullptr, "GetGyroscopeOdr"},
+            {151, nullptr, "SetGyroscopeOdr"},
+            {152, nullptr, "GetWhoAmI"},
             {201, nullptr, "ActivateFirmwareUpdate"},
             {202, nullptr, "DeactivateFirmwareUpdate"},
             {203, nullptr, "StartFirmwareUpdate"},
@@ -927,6 +951,17 @@ public:
             {233, nullptr, "ClearPairingInfo"},
             {234, nullptr, "GetUniquePadDeviceTypeSetInternal"},
             {235, nullptr, "EnableAnalogStickPower"},
+            {236, nullptr, "RequestKuinaUartClockCal"},
+            {237, nullptr, "GetKuinaUartClockCal"},
+            {238, nullptr, "SetKuinaUartClockTrim"},
+            {239, nullptr, "KuinaLoopbackTest"},
+            {240, nullptr, "RequestBatteryVoltage"},
+            {241, nullptr, "GetBatteryVoltage"},
+            {242, nullptr, "GetUniquePadPowerInfo"},
+            {243, nullptr, "RebootUniquePad"},
+            {244, nullptr, "RequestKuinaFirmwareVersion"},
+            {245, nullptr, "GetKuinaFirmwareVersion"},
+            {246, nullptr, "GetVidPid"},
             {301, nullptr, "GetAbstractedPadHandles"},
             {302, nullptr, "GetAbstractedPadState"},
             {303, nullptr, "GetAbstractedPadsState"},
@@ -945,6 +980,17 @@ public:
             {350, nullptr, "AddRegisteredDevice"},
             {400, nullptr, "DisableExternalMcuOnNxDevice"},
             {401, nullptr, "DisableRailDeviceFiltering"},
+            {402, nullptr, "EnableWiredPairing"},
+            {403, nullptr, "EnableShipmentModeAutoClear"},
+            {500, nullptr, "SetFactoryInt"},
+            {501, nullptr, "IsFactoryBootEnabled"},
+            {550, nullptr, "SetAnalogStickModelDataTemporarily"},
+            {551, nullptr, "GetAnalogStickModelData"},
+            {552, nullptr, "ResetAnalogStickModelData"},
+            {600, nullptr, "ConvertPadState"},
+            {2000, nullptr, "DeactivateDigitizer"},
+            {2001, nullptr, "SetDigitizerAutoPilotState"},
+            {2002, nullptr, "UnsetDigitizerAutoPilotState"},
         };
         // clang-format on
 
diff --git a/src/core/hle/service/hid/hid.h b/src/core/hle/service/hid/hid.h
index 039c38b58..23552efb1 100644
--- a/src/core/hle/service/hid/hid.h
+++ b/src/core/hle/service/hid/hid.h
@@ -128,6 +128,7 @@ private:
     void StopSixAxisSensor(Kernel::HLERequestContext& ctx);
     void SetIsPalmaAllConnectable(Kernel::HLERequestContext& ctx);
     void SetPalmaBoostMode(Kernel::HLERequestContext& ctx);
+    void StopSevenSixAxisSensor(Kernel::HLERequestContext& ctx);
     void InitializeSevenSixAxisSensor(Kernel::HLERequestContext& ctx);
 
     std::shared_ptr<IAppletResource> applet_resource;
diff --git a/src/core/hle/service/ldr/ldr.cpp b/src/core/hle/service/ldr/ldr.cpp
index 0cde7a557..6ad3be1b3 100644
--- a/src/core/hle/service/ldr/ldr.cpp
+++ b/src/core/hle/service/ldr/ldr.cpp
@@ -116,6 +116,7 @@ public:
             {1, nullptr, "GetProgramInfo"},
             {2, nullptr, "RegisterTitle"},
             {3, nullptr, "UnregisterTitle"},
+            {4, nullptr, "SetEnabledProgramVerification"},
         };
         // clang-format on
 
diff --git a/src/core/hle/service/ncm/ncm.cpp b/src/core/hle/service/ncm/ncm.cpp
index 89e283ca5..ec9aae04a 100644
--- a/src/core/hle/service/ncm/ncm.cpp
+++ b/src/core/hle/service/ncm/ncm.cpp
@@ -122,6 +122,7 @@ public:
             {11, nullptr, "ActivateContentMetaDatabase"},
             {12, nullptr, "InactivateContentMetaDatabase"},
             {13, nullptr, "InvalidateRightsIdCache"},
+            {14, nullptr, "GetMemoryReport"},
         };
         // clang-format on
 
diff --git a/src/core/hle/service/npns/npns.cpp b/src/core/hle/service/npns/npns.cpp
index aa171473b..f38d01084 100644
--- a/src/core/hle/service/npns/npns.cpp
+++ b/src/core/hle/service/npns/npns.cpp
@@ -48,6 +48,8 @@ public:
             {151, nullptr, "GetStateWithHandover"},
             {152, nullptr, "GetStateChangeEventWithHandover"},
             {153, nullptr, "GetDropEventWithHandover"},
+            {161, nullptr, "GetRequestChangeStateCancelEvent"},
+            {162, nullptr, "RequestChangeStateForceTimedWithCancelEvent"},
             {201, nullptr, "RequestChangeStateForceTimed"},
             {202, nullptr, "RequestChangeStateForceAsync"},
         };
diff --git a/src/core/hle/service/ns/ns.cpp b/src/core/hle/service/ns/ns.cpp
index fdab3cf78..8fb88990e 100644
--- a/src/core/hle/service/ns/ns.cpp
+++ b/src/core/hle/service/ns/ns.cpp
@@ -110,6 +110,10 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {100, nullptr, "ResetToFactorySettings"},
         {101, nullptr, "ResetToFactorySettingsWithoutUserSaveData"},
         {102, nullptr, "ResetToFactorySettingsForRefurbishment"},
+        {103, nullptr, "ResetToFactorySettingsWithPlatformRegion"},
+        {104, nullptr, "ResetToFactorySettingsWithPlatformRegionAuthentication"},
+        {105, nullptr, "RequestResetToFactorySettingsSecurely"},
+        {106, nullptr, "RequestResetToFactorySettingsWithPlatformRegionAuthenticationSecurely"},
         {200, nullptr, "CalculateUserSaveDataStatistics"},
         {201, nullptr, "DeleteUserSaveDataAll"},
         {210, nullptr, "DeleteUserSystemSaveData"},
@@ -191,6 +195,9 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {1307, nullptr, "TryDeleteRunningApplicationContentEntities"},
         {1308, nullptr, "DeleteApplicationCompletelyForDebug"},
         {1309, nullptr, "CleanupUnavailableAddOnContents"},
+        {1310, nullptr, "RequestMoveApplicationEntity"},
+        {1311, nullptr, "EstimateSizeToMove"},
+        {1312, nullptr, "HasMovableEntity"},
         {1400, nullptr, "PrepareShutdown"},
         {1500, nullptr, "FormatSdCard"},
         {1501, nullptr, "NeedsSystemUpdateToFormatSdCard"},
@@ -241,7 +248,7 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {2153, nullptr, "DeactivateRightsEnvironment"},
         {2154, nullptr, "ForceActivateRightsContextForExit"},
         {2155, nullptr, "UpdateRightsEnvironmentStatus"},
-        {2156, nullptr, "CreateRightsEnvironmentForPreomia"},
+        {2156, nullptr, "CreateRightsEnvironmentForMicroApplication"},
         {2160, nullptr, "AddTargetApplicationToRightsEnvironment"},
         {2161, nullptr, "SetUsersToRightsEnvironment"},
         {2170, nullptr, "GetRightsEnvironmentStatus"},
@@ -258,6 +265,7 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {2350, nullptr, "PerformAutoUpdateByApplicationId"},
         {2351, nullptr, "RequestNoDownloadRightsErrorResolution"},
         {2352, nullptr, "RequestResolveNoDownloadRightsError"},
+        {2353, nullptr, "GetApplicationDownloadTaskInfo"},
         {2400, nullptr, "GetPromotionInfo"},
         {2401, nullptr, "CountPromotionInfo"},
         {2402, nullptr, "ListPromotionInfo"},
@@ -266,9 +274,12 @@ IApplicationManagerInterface::IApplicationManagerInterface()
         {2500, nullptr, "ConfirmAvailableTime"},
         {2510, nullptr, "CreateApplicationResource"},
         {2511, nullptr, "GetApplicationResource"},
-        {2513, nullptr, "LaunchPreomia"},
+        {2513, nullptr, "LaunchMicroApplication"},
         {2514, nullptr, "ClearTaskOfAsyncTaskManager"},
+        {2515, nullptr, "CleanupAllPlaceHolderAndFragmentsIfNoTask"},
+        {2516, nullptr, "EnsureApplicationCertificate"},
         {2800, nullptr, "GetApplicationIdOfPreomia"},
+        {9999, nullptr, "GetApplicationCertificate"},
     };
     // clang-format on
 
@@ -505,6 +516,10 @@ IFactoryResetInterface::IFactoryResetInterface::IFactoryResetInterface()
             {100, nullptr, "ResetToFactorySettings"},
             {101, nullptr, "ResetToFactorySettingsWithoutUserSaveData"},
             {102, nullptr, "ResetToFactorySettingsForRefurbishment"},
+            {103, nullptr, "ResetToFactorySettingsWithPlatformRegion"},
+            {104, nullptr, "ResetToFactorySettingsWithPlatformRegionAuthentication"},
+            {105, nullptr, "RequestResetToFactorySettingsSecurely"},
+            {106, nullptr, "RequestResetToFactorySettingsWithPlatformRegionAuthenticationSecurely"},
         };
     // clang-format on
 
@@ -553,6 +568,9 @@ public:
             {10, nullptr, "TerminateApplication2"},
             {11, nullptr, "GetRunningApplicationProcessId"},
             {12, nullptr, "SetCurrentApplicationRightsEnvironmentCanBeActive"},
+            {13, nullptr, "CreateApplicationResourceForDevelop"},
+            {14, nullptr, "IsPreomiaForDevelop"},
+            {15, nullptr, "GetApplicationProgramIdFromHost"},
         };
         // clang-format on
 
diff --git a/src/core/hle/service/ns/pl_u.cpp b/src/core/hle/service/ns/pl_u.cpp
index ab1746d28..6efdf1606 100644
--- a/src/core/hle/service/ns/pl_u.cpp
+++ b/src/core/hle/service/ns/pl_u.cpp
@@ -164,6 +164,7 @@ PL_U::PL_U(Core::System& system)
         {6, nullptr, "GetSharedFontInOrderOfPriorityForSystem"},
         {100, nullptr, "RequestApplicationFunctionAuthorization"},
         {101, nullptr, "RequestApplicationFunctionAuthorizationForSystem"},
+        {102, nullptr, "RequestApplicationFunctionAuthorizationByApplicationId"},
         {1000, nullptr, "LoadNgWordDataForPlatformRegionChina"},
         {1001, nullptr, "GetNgWordDataSizeForPlatformRegionChina"},
     };
diff --git a/src/core/hle/service/pctl/module.cpp b/src/core/hle/service/pctl/module.cpp
index c75b4ee34..caf14ed61 100644
--- a/src/core/hle/service/pctl/module.cpp
+++ b/src/core/hle/service/pctl/module.cpp
@@ -31,6 +31,8 @@ public:
             {1014, nullptr, "ConfirmPlayableApplicationVideoOld"},
             {1015, nullptr, "ConfirmPlayableApplicationVideo"},
             {1016, nullptr, "ConfirmShowNewsPermission"},
+            {1017, nullptr, "EndFreeCommunication"},
+            {1018, nullptr, "IsFreeCommunicationAvailable"},
             {1031, nullptr, "IsRestrictionEnabled"},
             {1032, nullptr, "GetSafetyLevel"},
             {1033, nullptr, "SetSafetyLevel"},
diff --git a/src/core/hle/service/prepo/prepo.cpp b/src/core/hle/service/prepo/prepo.cpp
index 8f1be0e48..14309c679 100644
--- a/src/core/hle/service/prepo/prepo.cpp
+++ b/src/core/hle/service/prepo/prepo.cpp
@@ -21,8 +21,10 @@ public:
         static const FunctionInfo functions[] = {
             {10100, &PlayReport::SaveReport<Core::Reporter::PlayReportType::Old>, "SaveReportOld"},
             {10101, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::Old>, "SaveReportWithUserOld"},
-            {10102, &PlayReport::SaveReport<Core::Reporter::PlayReportType::New>, "SaveReport"},
-            {10103, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::New>, "SaveReportWithUser"},
+            {10102, &PlayReport::SaveReport<Core::Reporter::PlayReportType::Old2>, "SaveReportOld2"},
+            {10103, &PlayReport::SaveReportWithUser<Core::Reporter::PlayReportType::Old2>, "SaveReportWithUserOld2"},
+            {10104, nullptr, "SaveReport"},
+            {10105, nullptr, "SaveReportWithUser"},
             {10200, nullptr, "RequestImmediateTransmission"},
             {10300, nullptr, "GetTransmissionStatus"},
             {10400, nullptr, "GetSystemSessionId"},
@@ -35,8 +37,10 @@ public:
             {30400, nullptr, "GetStatistics"},
             {30401, nullptr, "GetThroughputHistory"},
             {30500, nullptr, "GetLastUploadError"},
+            {30600, nullptr, "GetApplicationUploadSummary"},
             {40100, nullptr, "IsUserAgreementCheckEnabled"},
             {40101, nullptr, "SetUserAgreementCheckEnabled"},
+            {50100, nullptr, "ReadAllApplicationReportFiles"},
             {90100, nullptr, "ReadAllReportFiles"},
         };
         // clang-format on
@@ -51,7 +55,7 @@ private:
         const auto process_id = rp.PopRaw<u64>();
 
         std::vector<std::vector<u8>> data{ctx.ReadBuffer(0)};
-        if (Type == Core::Reporter::PlayReportType::New) {
+        if constexpr (Type == Core::Reporter::PlayReportType::Old2) {
             data.emplace_back(ctx.ReadBuffer(1));
         }
 
@@ -71,7 +75,7 @@ private:
         const auto user_id = rp.PopRaw<u128>();
         const auto process_id = rp.PopRaw<u64>();
         std::vector<std::vector<u8>> data{ctx.ReadBuffer(0)};
-        if (Type == Core::Reporter::PlayReportType::New) {
+        if constexpr (Type == Core::Reporter::PlayReportType::Old2) {
             data.emplace_back(ctx.ReadBuffer(1));
         }
 
diff --git a/src/core/hle/service/set/set_cal.cpp b/src/core/hle/service/set/set_cal.cpp
index 1398a4a48..3fbfecc9e 100644
--- a/src/core/hle/service/set/set_cal.cpp
+++ b/src/core/hle/service/set/set_cal.cpp
@@ -50,6 +50,8 @@ SET_CAL::SET_CAL() : ServiceFramework("set:cal") {
         {39, nullptr, "GetConsoleSixAxisSensorModuleType"},
         {40, nullptr, "GetConsoleSixAxisSensorHorizontalOffset"},
         {41, nullptr, "GetBatteryVersion"},
+        {42, nullptr, "GetDeviceId"},
+        {43, nullptr, "GetConsoleSixAxisSensorMountType"},
     };
     // clang-format on
 
diff --git a/src/core/hle/service/set/set_sys.cpp b/src/core/hle/service/set/set_sys.cpp
index b7c9ea74b..8bd4c7e79 100644
--- a/src/core/hle/service/set/set_sys.cpp
+++ b/src/core/hle/service/set/set_sys.cpp
@@ -288,6 +288,18 @@ SET_SYS::SET_SYS() : ServiceFramework("set:sys") {
         {186, nullptr, "GetMemoryUsageRateFlag"},
         {187, nullptr, "GetTouchScreenMode"},
         {188, nullptr, "SetTouchScreenMode"},
+        {189, nullptr, "GetButtonConfigSettingsFull"},
+        {190, nullptr, "SetButtonConfigSettingsFull"},
+        {191, nullptr, "GetButtonConfigSettingsEmbedded"},
+        {192, nullptr, "SetButtonConfigSettingsEmbedded"},
+        {193, nullptr, "GetButtonConfigSettingsLeft"},
+        {194, nullptr, "SetButtonConfigSettingsLeft"},
+        {195, nullptr, "GetButtonConfigSettingsRight"},
+        {196, nullptr, "SetButtonConfigSettingsRight"},
+        {197, nullptr, "GetButtonConfigRegisteredSettingsEmbedded"},
+        {198, nullptr, "SetButtonConfigRegisteredSettingsEmbedded"},
+        {199, nullptr, "GetButtonConfigRegisteredSettings"},
+        {200, nullptr, "SetButtonConfigRegisteredSettings"},
     };
     // clang-format on
 
diff --git a/src/core/hle/service/sockets/bsd.cpp b/src/core/hle/service/sockets/bsd.cpp
index f67fab2f9..8d4952c0e 100644
--- a/src/core/hle/service/sockets/bsd.cpp
+++ b/src/core/hle/service/sockets/bsd.cpp
@@ -148,6 +148,7 @@ BSD::BSD(const char* name) : ServiceFramework(name) {
         {30, nullptr, "SendMMsg"},
         {31, nullptr, "EventFd"},
         {32, nullptr, "RegisterResourceStatisticsName"},
+        {33, nullptr, "Initialize2"},
     };
     // clang-format on
 
diff --git a/src/core/hle/service/time/time.cpp b/src/core/hle/service/time/time.cpp
index e722886de..67f1bbcf3 100644
--- a/src/core/hle/service/time/time.cpp
+++ b/src/core/hle/service/time/time.cpp
@@ -20,8 +20,8 @@ namespace Service::Time {
 
 class ISystemClock final : public ServiceFramework<ISystemClock> {
 public:
-    ISystemClock(Clock::SystemClockCore& clock_core)
-        : ServiceFramework("ISystemClock"), clock_core{clock_core} {
+    explicit ISystemClock(Clock::SystemClockCore& clock_core, Core::System& system)
+        : ServiceFramework("ISystemClock"), clock_core{clock_core}, system{system} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &ISystemClock::GetCurrentTime, "GetCurrentTime"},
@@ -46,9 +46,8 @@ private:
         }
 
         s64 posix_time{};
-        if (const ResultCode result{
-                clock_core.GetCurrentTime(Core::System::GetInstance(), posix_time)};
-            result != RESULT_SUCCESS) {
+        if (const ResultCode result{clock_core.GetCurrentTime(system, posix_time)};
+            result.IsError()) {
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(result);
             return;
@@ -69,9 +68,8 @@ private:
         }
 
         Clock::SystemClockContext system_clock_context{};
-        if (const ResultCode result{
-                clock_core.GetClockContext(Core::System::GetInstance(), system_clock_context)};
-            result != RESULT_SUCCESS) {
+        if (const ResultCode result{clock_core.GetClockContext(system, system_clock_context)};
+            result.IsError()) {
             IPC::ResponseBuilder rb{ctx, 2};
             rb.Push(result);
             return;
@@ -83,12 +81,13 @@ private:
     }
 
     Clock::SystemClockCore& clock_core;
+    Core::System& system;
 };
 
 class ISteadyClock final : public ServiceFramework<ISteadyClock> {
 public:
-    ISteadyClock(Clock::SteadyClockCore& clock_core)
-        : ServiceFramework("ISteadyClock"), clock_core{clock_core} {
+    explicit ISteadyClock(Clock::SteadyClockCore& clock_core, Core::System& system)
+        : ServiceFramework("ISteadyClock"), clock_core{clock_core}, system{system} {
         static const FunctionInfo functions[] = {
             {0, &ISteadyClock::GetCurrentTimePoint, "GetCurrentTimePoint"},
         };
@@ -105,14 +104,14 @@ private:
             return;
         }
 
-        const Clock::SteadyClockTimePoint time_point{
-            clock_core.GetCurrentTimePoint(Core::System::GetInstance())};
+        const Clock::SteadyClockTimePoint time_point{clock_core.GetCurrentTimePoint(system)};
         IPC::ResponseBuilder rb{ctx, (sizeof(Clock::SteadyClockTimePoint) / 4) + 2};
         rb.Push(RESULT_SUCCESS);
         rb.PushRaw(time_point);
     }
 
     Clock::SteadyClockCore& clock_core;
+    Core::System& system;
 };
 
 ResultCode Module::Interface::GetClockSnapshotFromSystemClockContextInternal(
@@ -134,7 +133,7 @@ ResultCode Module::Interface::GetClockSnapshotFromSystemClockContextInternal(
     }
 
     const auto current_time_point{
-        time_manager.GetStandardSteadyClockCore().GetCurrentTimePoint(Core::System::GetInstance())};
+        time_manager.GetStandardSteadyClockCore().GetCurrentTimePoint(system)};
     if (const ResultCode result{Clock::ClockSnapshot::GetCurrentTime(
             clock_snapshot.user_time, current_time_point, clock_snapshot.user_context)};
         result != RESULT_SUCCESS) {
@@ -176,21 +175,24 @@ void Module::Interface::GetStandardUserSystemClock(Kernel::HLERequestContext& ct
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardUserSystemClockCore());
+    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardUserSystemClockCore(),
+                                      system);
 }
 
 void Module::Interface::GetStandardNetworkSystemClock(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardNetworkSystemClockCore());
+    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardNetworkSystemClockCore(),
+                                      system);
 }
 
 void Module::Interface::GetStandardSteadyClock(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISteadyClock>(module->GetTimeManager().GetStandardSteadyClockCore());
+    rb.PushIpcInterface<ISteadyClock>(module->GetTimeManager().GetStandardSteadyClockCore(),
+                                      system);
 }
 
 void Module::Interface::GetTimeZoneService(Kernel::HLERequestContext& ctx) {
@@ -204,7 +206,8 @@ void Module::Interface::GetStandardLocalSystemClock(Kernel::HLERequestContext& c
     LOG_DEBUG(Service_Time, "called");
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(RESULT_SUCCESS);
-    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardLocalSystemClockCore());
+    rb.PushIpcInterface<ISystemClock>(module->GetTimeManager().GetStandardLocalSystemClockCore(),
+                                      system);
 }
 
 void Module::Interface::IsStandardNetworkSystemClockAccuracySufficient(
@@ -228,8 +231,7 @@ void Module::Interface::CalculateMonotonicSystemClockBaseTimePoint(Kernel::HLERe
 
     IPC::RequestParser rp{ctx};
     const auto context{rp.PopRaw<Clock::SystemClockContext>()};
-    const auto current_time_point{
-        steady_clock_core.GetCurrentTimePoint(Core::System::GetInstance())};
+    const auto current_time_point{steady_clock_core.GetCurrentTimePoint(system)};
 
     if (current_time_point.clock_source_id == context.steady_time_point.clock_source_id) {
         const auto ticks{Clock::TimeSpanType::FromTicks(
@@ -255,8 +257,8 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     Clock::SystemClockContext user_context{};
     if (const ResultCode result{
             module->GetTimeManager().GetStandardUserSystemClockCore().GetClockContext(
-                Core::System::GetInstance(), user_context)};
-        result != RESULT_SUCCESS) {
+                system, user_context)};
+        result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
         return;
@@ -264,8 +266,8 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     Clock::SystemClockContext network_context{};
     if (const ResultCode result{
             module->GetTimeManager().GetStandardNetworkSystemClockCore().GetClockContext(
-                Core::System::GetInstance(), network_context)};
-        result != RESULT_SUCCESS) {
+                system, network_context)};
+        result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
         return;
@@ -274,7 +276,7 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     Clock::ClockSnapshot clock_snapshot{};
     if (const ResultCode result{GetClockSnapshotFromSystemClockContextInternal(
             &ctx.GetThread(), user_context, network_context, type, clock_snapshot)};
-        result != RESULT_SUCCESS) {
+        result.IsError()) {
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
         return;
diff --git a/src/core/reporter.h b/src/core/reporter.h
index 380941b1b..86d760cf0 100644
--- a/src/core/reporter.h
+++ b/src/core/reporter.h
@@ -56,6 +56,7 @@ public:
 
     enum class PlayReportType {
         Old,
+        Old2,
         New,
         System,
     };
diff --git a/src/core/settings.cpp b/src/core/settings.cpp
index c1282cb80..cd6c257f5 100644
--- a/src/core/settings.cpp
+++ b/src/core/settings.cpp
@@ -92,7 +92,7 @@ void LogSettings() {
     LogSetting("Renderer_UseFrameLimit", Settings::values.use_frame_limit);
     LogSetting("Renderer_FrameLimit", Settings::values.frame_limit);
     LogSetting("Renderer_UseDiskShaderCache", Settings::values.use_disk_shader_cache);
-    LogSetting("Renderer_UseAccurateGpuEmulation", Settings::values.use_accurate_gpu_emulation);
+    LogSetting("Renderer_GPUAccuracyLevel", Settings::values.gpu_accuracy);
     LogSetting("Renderer_UseAsynchronousGpuEmulation",
                Settings::values.use_asynchronous_gpu_emulation);
     LogSetting("Renderer_UseVsync", Settings::values.use_vsync);
@@ -109,4 +109,12 @@ void LogSettings() {
     LogSetting("Services_BCATBoxcatLocal", Settings::values.bcat_boxcat_local);
 }
 
+bool IsGPULevelExtreme() {
+    return values.gpu_accuracy == GPUAccuracy::Extreme;
+}
+
+bool IsGPULevelHigh() {
+    return values.gpu_accuracy == GPUAccuracy::Extreme || values.gpu_accuracy == GPUAccuracy::High;
+}
+
 } // namespace Settings
diff --git a/src/core/settings.h b/src/core/settings.h
index c73d1c596..163900f0b 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -376,6 +376,12 @@ enum class RendererBackend {
     Vulkan = 1,
 };
 
+enum class GPUAccuracy : u32 {
+    Normal = 0,
+    High = 1,
+    Extreme = 2,
+};
+
 struct Values {
     // System
     bool use_docked_mode;
@@ -436,10 +442,11 @@ struct Values {
     bool use_frame_limit;
     u16 frame_limit;
     bool use_disk_shader_cache;
-    bool use_accurate_gpu_emulation;
+    GPUAccuracy gpu_accuracy;
     bool use_asynchronous_gpu_emulation;
     bool use_vsync;
     bool force_30fps_mode;
+    bool use_fast_gpu_time;
 
     float bg_red;
     float bg_green;
@@ -480,6 +487,9 @@ struct Values {
     std::map<u64, std::vector<std::string>> disabled_addons;
 } extern values;
 
+bool IsGPULevelExtreme();
+bool IsGPULevelHigh();
+
 void Apply();
 void LogSettings();
 } // namespace Settings
diff --git a/src/core/telemetry_session.cpp b/src/core/telemetry_session.cpp
index fd5a3ee9f..1c3b03a1c 100644
--- a/src/core/telemetry_session.cpp
+++ b/src/core/telemetry_session.cpp
@@ -56,6 +56,18 @@ static const char* TranslateRenderer(Settings::RendererBackend backend) {
     return "Unknown";
 }
 
+static const char* TranslateGPUAccuracyLevel(Settings::GPUAccuracy backend) {
+    switch (backend) {
+    case Settings::GPUAccuracy::Normal:
+        return "Normal";
+    case Settings::GPUAccuracy::High:
+        return "High";
+    case Settings::GPUAccuracy::Extreme:
+        return "Extreme";
+    }
+    return "Unknown";
+}
+
 u64 GetTelemetryId() {
     u64 telemetry_id{};
     const std::string filename{FileUtil::GetUserPath(FileUtil::UserPath::ConfigDir) +
@@ -184,8 +196,8 @@ void TelemetrySession::AddInitialInfo(Loader::AppLoader& app_loader) {
     AddField(field_type, "Renderer_UseFrameLimit", Settings::values.use_frame_limit);
     AddField(field_type, "Renderer_FrameLimit", Settings::values.frame_limit);
     AddField(field_type, "Renderer_UseDiskShaderCache", Settings::values.use_disk_shader_cache);
-    AddField(field_type, "Renderer_UseAccurateGpuEmulation",
-             Settings::values.use_accurate_gpu_emulation);
+    AddField(field_type, "Renderer_GPUAccuracyLevel",
+             TranslateGPUAccuracyLevel(Settings::values.gpu_accuracy));
     AddField(field_type, "Renderer_UseAsynchronousGpuEmulation",
              Settings::values.use_asynchronous_gpu_emulation);
     AddField(field_type, "Renderer_UseVsync", Settings::values.use_vsync);
diff --git a/src/tests/core/core_timing.cpp b/src/tests/core/core_timing.cpp
index 1e3940801..ff2d11cc8 100644
--- a/src/tests/core/core_timing.cpp
+++ b/src/tests/core/core_timing.cpp
@@ -14,13 +14,14 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 
+namespace {
 // Numbers are chosen randomly to make sure the correct one is given.
-static constexpr std::array<u64, 5> CB_IDS{{42, 144, 93, 1026, UINT64_C(0xFFFF7FFFF7FFFF)}};
-static constexpr int MAX_SLICE_LENGTH = 10000; // Copied from CoreTiming internals
+constexpr std::array<u64, 5> CB_IDS{{42, 144, 93, 1026, UINT64_C(0xFFFF7FFFF7FFFF)}};
+constexpr int MAX_SLICE_LENGTH = 10000; // Copied from CoreTiming internals
 
-static std::bitset<CB_IDS.size()> callbacks_ran_flags;
-static u64 expected_callback = 0;
-static s64 lateness = 0;
+std::bitset<CB_IDS.size()> callbacks_ran_flags;
+u64 expected_callback = 0;
+s64 lateness = 0;
 
 template <unsigned int IDX>
 void CallbackTemplate(u64 userdata, s64 cycles_late) {
@@ -31,7 +32,7 @@ void CallbackTemplate(u64 userdata, s64 cycles_late) {
     REQUIRE(lateness == cycles_late);
 }
 
-static u64 callbacks_done = 0;
+u64 callbacks_done = 0;
 
 void EmptyCallback(u64 userdata, s64 cycles_late) {
     ++callbacks_done;
@@ -48,8 +49,8 @@ struct ScopeInit final {
     Core::Timing::CoreTiming core_timing;
 };
 
-static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, u32 context = 0,
-                            int expected_lateness = 0, int cpu_downcount = 0) {
+void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, u32 context = 0,
+                     int expected_lateness = 0, int cpu_downcount = 0) {
     callbacks_ran_flags = 0;
     expected_callback = CB_IDS[idx];
     lateness = expected_lateness;
@@ -62,6 +63,7 @@ static void AdvanceAndCheck(Core::Timing::CoreTiming& core_timing, u32 idx, u32
 
     REQUIRE(decltype(callbacks_ran_flags)().set(idx) == callbacks_ran_flags);
 }
+} // Anonymous namespace
 
 TEST_CASE("CoreTiming[BasicOrder]", "[core]") {
     ScopeInit guard;
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 258d58eba..8ede4ba9b 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -23,6 +23,7 @@ add_library(video_core STATIC
     engines/shader_bytecode.h
     engines/shader_header.h
     engines/shader_type.h
+    fence_manager.h
     gpu.cpp
     gpu.h
     gpu_asynch.cpp
@@ -51,6 +52,8 @@ add_library(video_core STATIC
     renderer_opengl/gl_buffer_cache.h
     renderer_opengl/gl_device.cpp
     renderer_opengl/gl_device.h
+    renderer_opengl/gl_fence_manager.cpp
+    renderer_opengl/gl_fence_manager.h
     renderer_opengl/gl_framebuffer_cache.cpp
     renderer_opengl/gl_framebuffer_cache.h
     renderer_opengl/gl_rasterizer.cpp
@@ -160,6 +163,8 @@ if (ENABLE_VULKAN)
         renderer_vulkan/fixed_pipeline_state.h
         renderer_vulkan/maxwell_to_vk.cpp
         renderer_vulkan/maxwell_to_vk.h
+        renderer_vulkan/nsight_aftermath_tracker.cpp
+        renderer_vulkan/nsight_aftermath_tracker.h
         renderer_vulkan/renderer_vulkan.h
         renderer_vulkan/renderer_vulkan.cpp
         renderer_vulkan/vk_blit_screen.cpp
@@ -174,6 +179,8 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_descriptor_pool.h
         renderer_vulkan/vk_device.cpp
         renderer_vulkan/vk_device.h
+        renderer_vulkan/vk_fence_manager.cpp
+        renderer_vulkan/vk_fence_manager.h
         renderer_vulkan/vk_graphics_pipeline.cpp
         renderer_vulkan/vk_graphics_pipeline.h
         renderer_vulkan/vk_image.cpp
@@ -213,19 +220,30 @@ if (ENABLE_VULKAN)
         renderer_vulkan/wrapper.cpp
         renderer_vulkan/wrapper.h
     )
-
-    target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)
-    target_compile_definitions(video_core PRIVATE HAS_VULKAN)
 endif()
 
 create_target_directory_groups(video_core)
 
 target_link_libraries(video_core PUBLIC common core)
 target_link_libraries(video_core PRIVATE glad)
+
 if (ENABLE_VULKAN)
+    target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)
+    target_compile_definitions(video_core PRIVATE HAS_VULKAN)
     target_link_libraries(video_core PRIVATE sirit)
 endif()
 
+if (ENABLE_NSIGHT_AFTERMATH)
+    if (NOT DEFINED ENV{NSIGHT_AFTERMATH_SDK})
+        message(ERROR "Environment variable NSIGHT_AFTERMATH_SDK has to be provided")
+    endif()
+    if (NOT WIN32)
+        message(ERROR "Nsight Aftermath doesn't support non-Windows platforms")
+    endif()
+    target_compile_definitions(video_core PRIVATE HAS_NSIGHT_AFTERMATH)
+    target_include_directories(video_core PRIVATE "$ENV{NSIGHT_AFTERMATH_SDK}/include")
+endif()
+
 if (MSVC)
     target_compile_options(video_core PRIVATE /we4267)
 else()
diff --git a/src/video_core/buffer_cache/buffer_cache.h b/src/video_core/buffer_cache/buffer_cache.h
index 83e7a1cde..510f11089 100644
--- a/src/video_core/buffer_cache/buffer_cache.h
+++ b/src/video_core/buffer_cache/buffer_cache.h
@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <list>
 #include <memory>
 #include <mutex>
 #include <unordered_map>
@@ -18,8 +19,10 @@
 
 #include "common/alignment.h"
 #include "common/common_types.h"
+#include "common/logging/log.h"
 #include "core/core.h"
 #include "core/memory.h"
+#include "core/settings.h"
 #include "video_core/buffer_cache/buffer_block.h"
 #include "video_core/buffer_cache/map_interval.h"
 #include "video_core/memory_manager.h"
@@ -79,6 +82,9 @@ public:
         auto map = MapAddress(block, gpu_addr, cpu_addr, size);
         if (is_written) {
             map->MarkAsModified(true, GetModifiedTicks());
+            if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
+                MarkForAsyncFlush(map);
+            }
             if (!map->IsWritten()) {
                 map->MarkAsWritten(true);
                 MarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
@@ -137,11 +143,22 @@ public:
         });
         for (auto& object : objects) {
             if (object->IsModified() && object->IsRegistered()) {
+                mutex.unlock();
                 FlushMap(object);
+                mutex.lock();
             }
         }
     }
 
+    bool MustFlushRegion(VAddr addr, std::size_t size) {
+        std::lock_guard lock{mutex};
+
+        const std::vector<MapInterval> objects = GetMapsInRange(addr, size);
+        return std::any_of(objects.cbegin(), objects.cend(), [](const MapInterval& map) {
+            return map->IsModified() && map->IsRegistered();
+        });
+    }
+
     /// Mark the specified region as being invalidated
     void InvalidateRegion(VAddr addr, u64 size) {
         std::lock_guard lock{mutex};
@@ -154,6 +171,77 @@ public:
         }
     }
 
+    void OnCPUWrite(VAddr addr, std::size_t size) {
+        std::lock_guard lock{mutex};
+
+        for (const auto& object : GetMapsInRange(addr, size)) {
+            if (object->IsMemoryMarked() && object->IsRegistered()) {
+                UnmarkMemory(object);
+                object->SetSyncPending(true);
+                marked_for_unregister.emplace_back(object);
+            }
+        }
+    }
+
+    void SyncGuestHost() {
+        std::lock_guard lock{mutex};
+
+        for (const auto& object : marked_for_unregister) {
+            if (object->IsRegistered()) {
+                object->SetSyncPending(false);
+                Unregister(object);
+            }
+        }
+        marked_for_unregister.clear();
+    }
+
+    void CommitAsyncFlushes() {
+        if (uncommitted_flushes) {
+            auto commit_list = std::make_shared<std::list<MapInterval>>();
+            for (auto& map : *uncommitted_flushes) {
+                if (map->IsRegistered() && map->IsModified()) {
+                    // TODO(Blinkhawk): Implement backend asynchronous flushing
+                    // AsyncFlushMap(map)
+                    commit_list->push_back(map);
+                }
+            }
+            if (!commit_list->empty()) {
+                committed_flushes.push_back(commit_list);
+            } else {
+                committed_flushes.emplace_back();
+            }
+        } else {
+            committed_flushes.emplace_back();
+        }
+        uncommitted_flushes.reset();
+    }
+
+    bool ShouldWaitAsyncFlushes() const {
+        return !committed_flushes.empty() && committed_flushes.front() != nullptr;
+    }
+
+    bool HasUncommittedFlushes() const {
+        return uncommitted_flushes != nullptr;
+    }
+
+    void PopAsyncFlushes() {
+        if (committed_flushes.empty()) {
+            return;
+        }
+        auto& flush_list = committed_flushes.front();
+        if (!flush_list) {
+            committed_flushes.pop_front();
+            return;
+        }
+        for (MapInterval& map : *flush_list) {
+            if (map->IsRegistered()) {
+                // TODO(Blinkhawk): Replace this for reading the asynchronous flush
+                FlushMap(map);
+            }
+        }
+        committed_flushes.pop_front();
+    }
+
     virtual BufferType GetEmptyBuffer(std::size_t size) = 0;
 
 protected:
@@ -196,17 +284,30 @@ protected:
         const IntervalType interval{new_map->GetStart(), new_map->GetEnd()};
         mapped_addresses.insert({interval, new_map});
         rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
+        new_map->SetMemoryMarked(true);
         if (inherit_written) {
             MarkRegionAsWritten(new_map->GetStart(), new_map->GetEnd() - 1);
             new_map->MarkAsWritten(true);
         }
     }
 
-    /// Unregisters an object from the cache
-    void Unregister(MapInterval& map) {
+    void UnmarkMemory(const MapInterval& map) {
+        if (!map->IsMemoryMarked()) {
+            return;
+        }
         const std::size_t size = map->GetEnd() - map->GetStart();
         rasterizer.UpdatePagesCachedCount(map->GetStart(), size, -1);
+        map->SetMemoryMarked(false);
+    }
+
+    /// Unregisters an object from the cache
+    void Unregister(const MapInterval& map) {
+        UnmarkMemory(map);
         map->MarkAsRegistered(false);
+        if (map->IsSyncPending()) {
+            marked_for_unregister.remove(map);
+            map->SetSyncPending(false);
+        }
         if (map->IsWritten()) {
             UnmarkRegionAsWritten(map->GetStart(), map->GetEnd() - 1);
         }
@@ -264,6 +365,9 @@ private:
         MapInterval new_map = CreateMap(new_start, new_end, new_gpu_addr);
         if (modified_inheritance) {
             new_map->MarkAsModified(true, GetModifiedTicks());
+            if (Settings::IsGPULevelHigh() && Settings::values.use_asynchronous_gpu_emulation) {
+                MarkForAsyncFlush(new_map);
+            }
         }
         Register(new_map, write_inheritance);
         return new_map;
@@ -450,6 +554,13 @@ private:
         return false;
     }
 
+    void MarkForAsyncFlush(MapInterval& map) {
+        if (!uncommitted_flushes) {
+            uncommitted_flushes = std::make_shared<std::unordered_set<MapInterval>>();
+        }
+        uncommitted_flushes->insert(map);
+    }
+
     VideoCore::RasterizerInterface& rasterizer;
     Core::System& system;
 
@@ -479,6 +590,10 @@ private:
     u64 modified_ticks = 0;
 
     std::vector<u8> staging_buffer;
+    std::list<MapInterval> marked_for_unregister;
+
+    std::shared_ptr<std::unordered_set<MapInterval>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::list<MapInterval>>> committed_flushes;
 
     std::recursive_mutex mutex;
 };
diff --git a/src/video_core/buffer_cache/map_interval.h b/src/video_core/buffer_cache/map_interval.h
index b0956029d..29d8b26f3 100644
--- a/src/video_core/buffer_cache/map_interval.h
+++ b/src/video_core/buffer_cache/map_interval.h
@@ -46,6 +46,22 @@ public:
         return is_registered;
     }
 
+    void SetMemoryMarked(bool is_memory_marked_) {
+        is_memory_marked = is_memory_marked_;
+    }
+
+    bool IsMemoryMarked() const {
+        return is_memory_marked;
+    }
+
+    void SetSyncPending(bool is_sync_pending_) {
+        is_sync_pending = is_sync_pending_;
+    }
+
+    bool IsSyncPending() const {
+        return is_sync_pending;
+    }
+
     VAddr GetStart() const {
         return start;
     }
@@ -83,6 +99,8 @@ private:
     bool is_written{};
     bool is_modified{};
     bool is_registered{};
+    bool is_memory_marked{};
+    bool is_sync_pending{};
     u64 ticks{};
 };
 
diff --git a/src/video_core/dma_pusher.cpp b/src/video_core/dma_pusher.cpp
index 0b77afc71..16311f05e 100644
--- a/src/video_core/dma_pusher.cpp
+++ b/src/video_core/dma_pusher.cpp
@@ -21,6 +21,7 @@ MICROPROFILE_DEFINE(DispatchCalls, "GPU", "Execute command buffer", MP_RGB(128,
 void DmaPusher::DispatchCalls() {
     MICROPROFILE_SCOPE(DispatchCalls);
 
+    gpu.SyncGuestHost();
     // On entering GPU code, assume all memory may be touched by the ARM core.
     gpu.Maxwell3D().OnMemoryWrite();
 
@@ -32,6 +33,8 @@ void DmaPusher::DispatchCalls() {
         }
     }
     gpu.FlushCommands();
+    gpu.SyncGuestHost();
+    gpu.OnCommandListEnd();
 }
 
 bool DmaPusher::Step() {
@@ -68,16 +71,22 @@ bool DmaPusher::Step() {
     gpu.MemoryManager().ReadBlockUnsafe(dma_get, command_headers.data(),
                                         command_list_header.size * sizeof(u32));
 
-    for (const CommandHeader& command_header : command_headers) {
+    for (std::size_t index = 0; index < command_headers.size();) {
+        const CommandHeader& command_header = command_headers[index];
 
-        // now, see if we're in the middle of a command
-        if (dma_state.length_pending) {
-            // Second word of long non-inc methods command - method count
-            dma_state.length_pending = 0;
-            dma_state.method_count = command_header.method_count_;
-        } else if (dma_state.method_count) {
+        if (dma_state.method_count) {
             // Data word of methods command
-            CallMethod(command_header.argument);
+            if (dma_state.non_incrementing) {
+                const u32 max_write = static_cast<u32>(
+                    std::min<std::size_t>(index + dma_state.method_count, command_headers.size()) -
+                    index);
+                CallMultiMethod(&command_header.argument, max_write);
+                dma_state.method_count -= max_write;
+                index += max_write;
+                continue;
+            } else {
+                CallMethod(command_header.argument);
+            }
 
             if (!dma_state.non_incrementing) {
                 dma_state.method++;
@@ -117,6 +126,7 @@ bool DmaPusher::Step() {
                 break;
             }
         }
+        index++;
     }
 
     if (!non_main) {
@@ -137,4 +147,9 @@ void DmaPusher::CallMethod(u32 argument) const {
     gpu.CallMethod({dma_state.method, argument, dma_state.subchannel, dma_state.method_count});
 }
 
+void DmaPusher::CallMultiMethod(const u32* base_start, u32 num_methods) const {
+    gpu.CallMultiMethod(dma_state.method, dma_state.subchannel, base_start, num_methods,
+                        dma_state.method_count);
+}
+
 } // namespace Tegra
diff --git a/src/video_core/dma_pusher.h b/src/video_core/dma_pusher.h
index d6188614a..6cef71306 100644
--- a/src/video_core/dma_pusher.h
+++ b/src/video_core/dma_pusher.h
@@ -75,6 +75,7 @@ private:
     void SetState(const CommandHeader& command_header);
 
     void CallMethod(u32 argument) const;
+    void CallMultiMethod(const u32* base_start, u32 num_methods) const;
 
     std::vector<CommandHeader> command_headers; ///< Buffer for list of commands fetched at once
 
diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp
index 85d308e26..8a47614d2 100644
--- a/src/video_core/engines/fermi_2d.cpp
+++ b/src/video_core/engines/fermi_2d.cpp
@@ -28,7 +28,13 @@ void Fermi2D::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
-std::pair<u32, u32> DelimitLine(u32 src_1, u32 src_2, u32 dst_1, u32 dst_2, u32 src_line) {
+void Fermi2D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
+static std::pair<u32, u32> DelimitLine(u32 src_1, u32 src_2, u32 dst_1, u32 dst_2, u32 src_line) {
     const u32 line_a = src_2 - src_1;
     const u32 line_b = dst_2 - dst_1;
     const u32 excess = std::max<s32>(0, line_a - src_line + src_1);
diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h
index dba342c70..939a5966d 100644
--- a/src/video_core/engines/fermi_2d.h
+++ b/src/video_core/engines/fermi_2d.h
@@ -39,6 +39,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     enum class Origin : u32 {
         Center = 0,
         Corner = 1,
diff --git a/src/video_core/engines/kepler_compute.cpp b/src/video_core/engines/kepler_compute.cpp
index 368c75a66..00a12175f 100644
--- a/src/video_core/engines/kepler_compute.cpp
+++ b/src/video_core/engines/kepler_compute.cpp
@@ -51,6 +51,13 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+void KeplerCompute::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                    u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
 Texture::FullTextureInfo KeplerCompute::GetTexture(std::size_t offset) const {
     const std::bitset<8> cbuf_mask = launch_description.const_buffer_enable_mask.Value();
     ASSERT(cbuf_mask[regs.tex_cb_index]);
diff --git a/src/video_core/engines/kepler_compute.h b/src/video_core/engines/kepler_compute.h
index eeb79c56f..fe55fdfd0 100644
--- a/src/video_core/engines/kepler_compute.h
+++ b/src/video_core/engines/kepler_compute.h
@@ -202,6 +202,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     Texture::FullTextureInfo GetTexture(std::size_t offset) const;
 
     /// Given a texture handle, returns the TSC and TIC entries.
diff --git a/src/video_core/engines/kepler_memory.cpp b/src/video_core/engines/kepler_memory.cpp
index 597872e43..586ff15dc 100644
--- a/src/video_core/engines/kepler_memory.cpp
+++ b/src/video_core/engines/kepler_memory.cpp
@@ -41,4 +41,11 @@ void KeplerMemory::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+void KeplerMemory::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                   u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
 } // namespace Tegra::Engines
diff --git a/src/video_core/engines/kepler_memory.h b/src/video_core/engines/kepler_memory.h
index 396fb6e86..bb26fb030 100644
--- a/src/video_core/engines/kepler_memory.h
+++ b/src/video_core/engines/kepler_memory.h
@@ -40,6 +40,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     struct Regs {
         static constexpr size_t NUM_REGS = 0x7F;
 
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index baa74ad4c..39e3b66a2 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -280,6 +280,58 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                u32 methods_pending) {
+    // Methods after 0xE00 are special, they're actually triggers for some microcode that was
+    // uploaded to the GPU during initialization.
+    if (method >= MacroRegistersStart) {
+        // We're trying to execute a macro
+        if (executing_macro == 0) {
+            // A macro call must begin by writing the macro method's register, not its argument.
+            ASSERT_MSG((method % 2) == 0,
+                       "Can't start macro execution by writing to the ARGS register");
+            executing_macro = method;
+        }
+
+        for (std::size_t i = 0; i < amount; i++) {
+            macro_params.push_back(base_start[i]);
+        }
+
+        // Call the macro when there are no more parameters in the command buffer
+        if (amount == methods_pending) {
+            CallMacroMethod(executing_macro, macro_params.size(), macro_params.data());
+            macro_params.clear();
+        }
+        return;
+    }
+    switch (method) {
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[1]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[2]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[3]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[4]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[5]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[6]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[7]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[8]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[9]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[10]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[11]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[12]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[13]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[14]):
+    case MAXWELL3D_REG_INDEX(const_buffer.cb_data[15]): {
+        ProcessCBMultiData(method, base_start, amount);
+        break;
+    }
+    default: {
+        for (std::size_t i = 0; i < amount; i++) {
+            CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+        }
+    }
+    }
+}
+
 void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) {
     if (mme_draw.current_mode == MMEDrawMode::Undefined) {
         if (mme_draw.gl_begin_consume) {
@@ -404,7 +456,11 @@ void Maxwell3D::ProcessQueryGet() {
 
     switch (regs.query.query_get.operation) {
     case Regs::QueryOperation::Release:
-        StampQueryResult(regs.query.query_sequence, regs.query.query_get.short_query == 0);
+        if (regs.query.query_get.fence == 1) {
+            rasterizer.SignalSemaphore(regs.query.QueryAddress(), regs.query.query_sequence);
+        } else {
+            StampQueryResult(regs.query.query_sequence, regs.query.query_get.short_query == 0);
+        }
         break;
     case Regs::QueryOperation::Acquire:
         // TODO(Blinkhawk): Under this operation, the GPU waits for the CPU to write a value that
@@ -483,7 +539,7 @@ void Maxwell3D::ProcessSyncPoint() {
     const u32 increment = regs.sync_info.increment.Value();
     [[maybe_unused]] const u32 cache_flush = regs.sync_info.unknown.Value();
     if (increment) {
-        system.GPU().IncrementSyncPoint(sync_point);
+        rasterizer.SignalSyncPoint(sync_point);
     }
 }
 
@@ -566,6 +622,28 @@ void Maxwell3D::StartCBData(u32 method) {
     ProcessCBData(regs.const_buffer.cb_data[cb_data_state.id]);
 }
 
+void Maxwell3D::ProcessCBMultiData(u32 method, const u32* start_base, u32 amount) {
+    if (cb_data_state.current != method) {
+        if (cb_data_state.current != null_cb_data) {
+            FinishCBData();
+        }
+        constexpr u32 first_cb_data = MAXWELL3D_REG_INDEX(const_buffer.cb_data[0]);
+        cb_data_state.start_pos = regs.const_buffer.cb_pos;
+        cb_data_state.id = method - first_cb_data;
+        cb_data_state.current = method;
+        cb_data_state.counter = 0;
+    }
+    const std::size_t id = cb_data_state.id;
+    const std::size_t size = amount;
+    std::size_t i = 0;
+    for (; i < size; i++) {
+        cb_data_state.buffer[id][cb_data_state.counter] = start_base[i];
+        cb_data_state.counter++;
+    }
+    // Increment the current buffer position.
+    regs.const_buffer.cb_pos = regs.const_buffer.cb_pos + 4 * amount;
+}
+
 void Maxwell3D::FinishCBData() {
     // Write the input value to the current const buffer at the current position.
     const GPUVAddr buffer_address = regs.const_buffer.BufferAddress();
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 59d5752d2..3dfba8197 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -1259,7 +1259,8 @@ public:
 
                     GPUVAddr LimitAddress() const {
                         return static_cast<GPUVAddr>((static_cast<GPUVAddr>(limit_high) << 32) |
-                                                     limit_low);
+                                                     limit_low) +
+                               1;
                     }
                 } vertex_array_limit[NumVertexArrays];
 
@@ -1358,6 +1359,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     /// Write the value to the register identified by method.
     void CallMethodFromMME(const GPU::MethodCall& method_call);
 
@@ -1511,6 +1515,7 @@ private:
     /// Handles a write to the CB_DATA[i] register.
     void StartCBData(u32 method);
     void ProcessCBData(u32 value);
+    void ProcessCBMultiData(u32 method, const u32* start_base, u32 amount);
     void FinishCBData();
 
     /// Handles a write to the CB_BIND register.
diff --git a/src/video_core/engines/maxwell_dma.cpp b/src/video_core/engines/maxwell_dma.cpp
index c2610f992..6630005b0 100644
--- a/src/video_core/engines/maxwell_dma.cpp
+++ b/src/video_core/engines/maxwell_dma.cpp
@@ -36,6 +36,13 @@ void MaxwellDMA::CallMethod(const GPU::MethodCall& method_call) {
 #undef MAXWELLDMA_REG_INDEX
 }
 
+void MaxwellDMA::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
+                                 u32 methods_pending) {
+    for (std::size_t i = 0; i < amount; i++) {
+        CallMethod({method, base_start[i], 0, methods_pending - static_cast<u32>(i)});
+    }
+}
+
 void MaxwellDMA::HandleCopy() {
     LOG_TRACE(HW_GPU, "Requested a DMA copy");
 
@@ -104,8 +111,13 @@ void MaxwellDMA::HandleCopy() {
             write_buffer.resize(dst_size);
         }
 
-        memory_manager.ReadBlock(source, read_buffer.data(), src_size);
-        memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
+        if (Settings::IsGPULevelExtreme()) {
+            memory_manager.ReadBlock(source, read_buffer.data(), src_size);
+            memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
+        } else {
+            memory_manager.ReadBlockUnsafe(source, read_buffer.data(), src_size);
+            memory_manager.ReadBlockUnsafe(dest, write_buffer.data(), dst_size);
+        }
 
         Texture::UnswizzleSubrect(
             regs.x_count, regs.y_count, regs.dst_pitch, regs.src_params.size_x, bytes_per_pixel,
@@ -136,7 +148,7 @@ void MaxwellDMA::HandleCopy() {
             write_buffer.resize(dst_size);
         }
 
-        if (Settings::values.use_accurate_gpu_emulation) {
+        if (Settings::IsGPULevelExtreme()) {
             memory_manager.ReadBlock(source, read_buffer.data(), src_size);
             memory_manager.ReadBlock(dest, write_buffer.data(), dst_size);
         } else {
diff --git a/src/video_core/engines/maxwell_dma.h b/src/video_core/engines/maxwell_dma.h
index 4f40d1d1f..c43ed8194 100644
--- a/src/video_core/engines/maxwell_dma.h
+++ b/src/video_core/engines/maxwell_dma.h
@@ -35,6 +35,9 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    /// Write multiple values to the register identified by method.
+    void CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32 methods_pending);
+
     struct Regs {
         static constexpr std::size_t NUM_REGS = 0x1D6;
 
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 7231597d4..cde3a26b9 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -655,6 +655,7 @@ union Instruction {
     }
 
     constexpr Instruction(u64 value) : value{value} {}
+    constexpr Instruction(const Instruction& instr) : value(instr.value) {}
 
     BitField<0, 8, Register> gpr0;
     BitField<8, 8, Register> gpr8;
@@ -817,11 +818,9 @@ union Instruction {
         BitField<32, 1, u64> saturate;
         BitField<49, 2, HalfMerge> merge;
 
-        BitField<43, 1, u64> negate_a;
         BitField<44, 1, u64> abs_a;
         BitField<47, 2, HalfType> type_a;
 
-        BitField<31, 1, u64> negate_b;
         BitField<30, 1, u64> abs_b;
         BitField<28, 2, HalfType> type_b;
 
diff --git a/src/video_core/fence_manager.h b/src/video_core/fence_manager.h
new file mode 100644
index 000000000..dabd1588c
--- /dev/null
+++ b/src/video_core/fence_manager.h
@@ -0,0 +1,170 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <array>
+#include <memory>
+#include <queue>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "core/core.h"
+#include "core/memory.h"
+#include "core/settings.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
+
+namespace VideoCommon {
+
+class FenceBase {
+public:
+    FenceBase(u32 payload, bool is_stubbed)
+        : address{}, payload{payload}, is_semaphore{false}, is_stubbed{is_stubbed} {}
+
+    FenceBase(GPUVAddr address, u32 payload, bool is_stubbed)
+        : address{address}, payload{payload}, is_semaphore{true}, is_stubbed{is_stubbed} {}
+
+    GPUVAddr GetAddress() const {
+        return address;
+    }
+
+    u32 GetPayload() const {
+        return payload;
+    }
+
+    bool IsSemaphore() const {
+        return is_semaphore;
+    }
+
+private:
+    GPUVAddr address;
+    u32 payload;
+    bool is_semaphore;
+
+protected:
+    bool is_stubbed;
+};
+
+template <typename TFence, typename TTextureCache, typename TTBufferCache, typename TQueryCache>
+class FenceManager {
+public:
+    void SignalSemaphore(GPUVAddr addr, u32 value) {
+        TryReleasePendingFences();
+        const bool should_flush = ShouldFlush();
+        CommitAsyncFlushes();
+        TFence new_fence = CreateFence(addr, value, !should_flush);
+        fences.push(new_fence);
+        QueueFence(new_fence);
+        if (should_flush) {
+            rasterizer.FlushCommands();
+        }
+        rasterizer.SyncGuestHost();
+    }
+
+    void SignalSyncPoint(u32 value) {
+        TryReleasePendingFences();
+        const bool should_flush = ShouldFlush();
+        CommitAsyncFlushes();
+        TFence new_fence = CreateFence(value, !should_flush);
+        fences.push(new_fence);
+        QueueFence(new_fence);
+        if (should_flush) {
+            rasterizer.FlushCommands();
+        }
+        rasterizer.SyncGuestHost();
+    }
+
+    void WaitPendingFences() {
+        auto& gpu{system.GPU()};
+        auto& memory_manager{gpu.MemoryManager()};
+        while (!fences.empty()) {
+            TFence& current_fence = fences.front();
+            if (ShouldWait()) {
+                WaitFence(current_fence);
+            }
+            PopAsyncFlushes();
+            if (current_fence->IsSemaphore()) {
+                memory_manager.Write<u32>(current_fence->GetAddress(), current_fence->GetPayload());
+            } else {
+                gpu.IncrementSyncPoint(current_fence->GetPayload());
+            }
+            fences.pop();
+        }
+    }
+
+protected:
+    FenceManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                 TTextureCache& texture_cache, TTBufferCache& buffer_cache,
+                 TQueryCache& query_cache)
+        : system{system}, rasterizer{rasterizer}, texture_cache{texture_cache},
+          buffer_cache{buffer_cache}, query_cache{query_cache} {}
+
+    virtual ~FenceManager() {}
+
+    /// Creates a Sync Point Fence Interface, does not create a backend fence if 'is_stubbed' is
+    /// true
+    virtual TFence CreateFence(u32 value, bool is_stubbed) = 0;
+    /// Creates a Semaphore Fence Interface, does not create a backend fence if 'is_stubbed' is true
+    virtual TFence CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) = 0;
+    /// Queues a fence into the backend if the fence isn't stubbed.
+    virtual void QueueFence(TFence& fence) = 0;
+    /// Notifies that the backend fence has been signaled/reached in host GPU.
+    virtual bool IsFenceSignaled(TFence& fence) const = 0;
+    /// Waits until a fence has been signalled by the host GPU.
+    virtual void WaitFence(TFence& fence) = 0;
+
+    Core::System& system;
+    VideoCore::RasterizerInterface& rasterizer;
+    TTextureCache& texture_cache;
+    TTBufferCache& buffer_cache;
+    TQueryCache& query_cache;
+
+private:
+    void TryReleasePendingFences() {
+        auto& gpu{system.GPU()};
+        auto& memory_manager{gpu.MemoryManager()};
+        while (!fences.empty()) {
+            TFence& current_fence = fences.front();
+            if (ShouldWait() && !IsFenceSignaled(current_fence)) {
+                return;
+            }
+            PopAsyncFlushes();
+            if (current_fence->IsSemaphore()) {
+                memory_manager.Write<u32>(current_fence->GetAddress(), current_fence->GetPayload());
+            } else {
+                gpu.IncrementSyncPoint(current_fence->GetPayload());
+            }
+            fences.pop();
+        }
+    }
+
+    bool ShouldWait() const {
+        return texture_cache.ShouldWaitAsyncFlushes() || buffer_cache.ShouldWaitAsyncFlushes() ||
+               query_cache.ShouldWaitAsyncFlushes();
+    }
+
+    bool ShouldFlush() const {
+        return texture_cache.HasUncommittedFlushes() || buffer_cache.HasUncommittedFlushes() ||
+               query_cache.HasUncommittedFlushes();
+    }
+
+    void PopAsyncFlushes() {
+        texture_cache.PopAsyncFlushes();
+        buffer_cache.PopAsyncFlushes();
+        query_cache.PopAsyncFlushes();
+    }
+
+    void CommitAsyncFlushes() {
+        texture_cache.CommitAsyncFlushes();
+        buffer_cache.CommitAsyncFlushes();
+        query_cache.CommitAsyncFlushes();
+    }
+
+    std::queue<TFence> fences;
+};
+
+} // namespace VideoCommon
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index a606f4abd..b87fd873d 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -9,6 +9,7 @@
 #include "core/core_timing_util.h"
 #include "core/frontend/emu_window.h"
 #include "core/memory.h"
+#include "core/settings.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/kepler_memory.h"
@@ -125,6 +126,28 @@ bool GPU::CancelSyncptInterrupt(const u32 syncpoint_id, const u32 value) {
     return true;
 }
 
+u64 GPU::RequestFlush(VAddr addr, std::size_t size) {
+    std::unique_lock lck{flush_request_mutex};
+    const u64 fence = ++last_flush_fence;
+    flush_requests.emplace_back(fence, addr, size);
+    return fence;
+}
+
+void GPU::TickWork() {
+    std::unique_lock lck{flush_request_mutex};
+    while (!flush_requests.empty()) {
+        auto& request = flush_requests.front();
+        const u64 fence = request.fence;
+        const VAddr addr = request.addr;
+        const std::size_t size = request.size;
+        flush_requests.pop_front();
+        flush_request_mutex.unlock();
+        renderer->Rasterizer().FlushRegion(addr, size);
+        current_flush_fence.store(fence);
+        flush_request_mutex.lock();
+    }
+}
+
 u64 GPU::GetTicks() const {
     // This values were reversed engineered by fincs from NVN
     // The gpu clock is reported in units of 385/625 nanoseconds
@@ -132,7 +155,10 @@ u64 GPU::GetTicks() const {
     constexpr u64 gpu_ticks_den = 625;
 
     const u64 cpu_ticks = system.CoreTiming().GetTicks();
-    const u64 nanoseconds = Core::Timing::CyclesToNs(cpu_ticks).count();
+    u64 nanoseconds = Core::Timing::CyclesToNs(cpu_ticks).count();
+    if (Settings::values.use_fast_gpu_time) {
+        nanoseconds /= 256;
+    }
     const u64 nanoseconds_num = nanoseconds / gpu_ticks_den;
     const u64 nanoseconds_rem = nanoseconds % gpu_ticks_den;
     return nanoseconds_num * gpu_ticks_num + (nanoseconds_rem * gpu_ticks_num) / gpu_ticks_den;
@@ -142,6 +168,13 @@ void GPU::FlushCommands() {
     renderer->Rasterizer().FlushCommands();
 }
 
+void GPU::SyncGuestHost() {
+    renderer->Rasterizer().SyncGuestHost();
+}
+
+void GPU::OnCommandListEnd() {
+    renderer->Rasterizer().ReleaseFences();
+}
 // Note that, traditionally, methods are treated as 4-byte addressable locations, and hence
 // their numbers are written down multiplied by 4 in Docs. Here we are not multiply by 4.
 // So the values you see in docs might be multiplied by 4.
@@ -180,16 +213,32 @@ void GPU::CallMethod(const MethodCall& method_call) {
 
     ASSERT(method_call.subchannel < bound_engines.size());
 
-    if (ExecuteMethodOnEngine(method_call)) {
+    if (ExecuteMethodOnEngine(method_call.method)) {
         CallEngineMethod(method_call);
     } else {
         CallPullerMethod(method_call);
     }
 }
 
-bool GPU::ExecuteMethodOnEngine(const MethodCall& method_call) {
-    const auto method = static_cast<BufferMethods>(method_call.method);
-    return method >= BufferMethods::NonPullerMethods;
+void GPU::CallMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                          u32 methods_pending) {
+    LOG_TRACE(HW_GPU, "Processing method {:08X} on subchannel {}", method, subchannel);
+
+    ASSERT(subchannel < bound_engines.size());
+
+    if (ExecuteMethodOnEngine(method)) {
+        CallEngineMultiMethod(method, subchannel, base_start, amount, methods_pending);
+    } else {
+        for (std::size_t i = 0; i < amount; i++) {
+            CallPullerMethod(
+                {method, base_start[i], subchannel, methods_pending - static_cast<u32>(i)});
+        }
+    }
+}
+
+bool GPU::ExecuteMethodOnEngine(u32 method) {
+    const auto buffer_method = static_cast<BufferMethods>(method);
+    return buffer_method >= BufferMethods::NonPullerMethods;
 }
 
 void GPU::CallPullerMethod(const MethodCall& method_call) {
@@ -269,6 +318,31 @@ void GPU::CallEngineMethod(const MethodCall& method_call) {
     }
 }
 
+void GPU::CallEngineMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                                u32 methods_pending) {
+    const EngineID engine = bound_engines[subchannel];
+
+    switch (engine) {
+    case EngineID::FERMI_TWOD_A:
+        fermi_2d->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::MAXWELL_B:
+        maxwell_3d->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::KEPLER_COMPUTE_B:
+        kepler_compute->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::MAXWELL_DMA_COPY_A:
+        maxwell_dma->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    case EngineID::KEPLER_INLINE_TO_MEMORY_B:
+        kepler_memory->CallMultiMethod(method, base_start, amount, methods_pending);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented engine");
+    }
+}
+
 void GPU::ProcessBindMethod(const MethodCall& method_call) {
     // Bind the current subchannel to the desired engine id.
     LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", method_call.subchannel,
diff --git a/src/video_core/gpu.h b/src/video_core/gpu.h
index 1a2d747be..dd51c95b7 100644
--- a/src/video_core/gpu.h
+++ b/src/video_core/gpu.h
@@ -155,7 +155,27 @@ public:
     /// Calls a GPU method.
     void CallMethod(const MethodCall& method_call);
 
+    /// Calls a GPU multivalue method.
+    void CallMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                         u32 methods_pending);
+
+    /// Flush all current written commands into the host GPU for execution.
     void FlushCommands();
+    /// Synchronizes CPU writes with Host GPU memory.
+    void SyncGuestHost();
+    /// Signal the ending of command list.
+    virtual void OnCommandListEnd();
+
+    /// Request a host GPU memory flush from the CPU.
+    u64 RequestFlush(VAddr addr, std::size_t size);
+
+    /// Obtains current flush request fence id.
+    u64 CurrentFlushRequestFence() const {
+        return current_flush_fence.load(std::memory_order_relaxed);
+    }
+
+    /// Tick pending requests within the GPU.
+    void TickWork();
 
     /// Returns a reference to the Maxwell3D GPU engine.
     Engines::Maxwell3D& Maxwell3D();
@@ -293,8 +313,12 @@ private:
     /// Calls a GPU engine method.
     void CallEngineMethod(const MethodCall& method_call);
 
+    /// Calls a GPU engine multivalue method.
+    void CallEngineMultiMethod(u32 method, u32 subchannel, const u32* base_start, u32 amount,
+                               u32 methods_pending);
+
     /// Determines where the method should be executed.
-    bool ExecuteMethodOnEngine(const MethodCall& method_call);
+    bool ExecuteMethodOnEngine(u32 method);
 
 protected:
     std::unique_ptr<Tegra::DmaPusher> dma_pusher;
@@ -325,6 +349,19 @@ private:
 
     std::condition_variable sync_cv;
 
+    struct FlushRequest {
+        FlushRequest(u64 fence, VAddr addr, std::size_t size)
+            : fence{fence}, addr{addr}, size{size} {}
+        u64 fence;
+        VAddr addr;
+        std::size_t size;
+    };
+
+    std::list<FlushRequest> flush_requests;
+    std::atomic<u64> current_flush_fence{};
+    u64 last_flush_fence{};
+    std::mutex flush_request_mutex;
+
     const bool is_async;
 };
 
diff --git a/src/video_core/gpu_asynch.cpp b/src/video_core/gpu_asynch.cpp
index 20e73a37e..53305ab43 100644
--- a/src/video_core/gpu_asynch.cpp
+++ b/src/video_core/gpu_asynch.cpp
@@ -52,4 +52,8 @@ 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
index 03fd0eef0..517658612 100644
--- a/src/video_core/gpu_asynch.h
+++ b/src/video_core/gpu_asynch.h
@@ -32,6 +32,8 @@ public:
     void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void WaitIdle() const override;
 
+    void OnCommandListEnd() override;
+
 protected:
     void TriggerCpuInterrupt(u32 syncpoint_id, u32 value) const override;
 
diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp
index 10cda686b..c3bb4fe06 100644
--- a/src/video_core/gpu_thread.cpp
+++ b/src/video_core/gpu_thread.cpp
@@ -6,6 +6,7 @@
 #include "common/microprofile.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
+#include "core/settings.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/gpu.h"
 #include "video_core/gpu_thread.h"
@@ -14,8 +15,9 @@
 namespace VideoCommon::GPUThread {
 
 /// Runs the GPU thread
-static void RunThread(VideoCore::RendererBase& renderer, Core::Frontend::GraphicsContext& context,
-                      Tegra::DmaPusher& dma_pusher, SynchState& state) {
+static void RunThread(Core::System& system, VideoCore::RendererBase& renderer,
+                      Core::Frontend::GraphicsContext& context, Tegra::DmaPusher& dma_pusher,
+                      SynchState& state) {
     MicroProfileOnThreadCreate("GpuThread");
 
     // Wait for first GPU command before acquiring the window context
@@ -37,10 +39,14 @@ static void RunThread(VideoCore::RendererBase& renderer, Core::Frontend::Graphic
             dma_pusher.DispatchCalls();
         } else if (const auto data = std::get_if<SwapBuffersCommand>(&next.data)) {
             renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
+        } else if (const auto data = std::get_if<OnCommandListEndCommand>(&next.data)) {
+            renderer.Rasterizer().ReleaseFences();
+        } else if (const auto data = std::get_if<GPUTickCommand>(&next.data)) {
+            system.GPU().TickWork();
         } else if (const auto data = std::get_if<FlushRegionCommand>(&next.data)) {
             renderer.Rasterizer().FlushRegion(data->addr, data->size);
         } else if (const auto data = std::get_if<InvalidateRegionCommand>(&next.data)) {
-            renderer.Rasterizer().InvalidateRegion(data->addr, data->size);
+            renderer.Rasterizer().OnCPUWrite(data->addr, data->size);
         } else if (std::holds_alternative<EndProcessingCommand>(next.data)) {
             return;
         } else {
@@ -65,8 +71,8 @@ ThreadManager::~ThreadManager() {
 void ThreadManager::StartThread(VideoCore::RendererBase& renderer,
                                 Core::Frontend::GraphicsContext& context,
                                 Tegra::DmaPusher& dma_pusher) {
-    thread = std::thread{RunThread, std::ref(renderer), std::ref(context), std::ref(dma_pusher),
-                         std::ref(state)};
+    thread = std::thread{RunThread,         std::ref(system),     std::ref(renderer),
+                         std::ref(context), std::ref(dma_pusher), std::ref(state)};
 }
 
 void ThreadManager::SubmitList(Tegra::CommandList&& entries) {
@@ -78,16 +84,29 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
 }
 
 void ThreadManager::FlushRegion(VAddr addr, u64 size) {
-    PushCommand(FlushRegionCommand(addr, size));
+    if (!Settings::IsGPULevelHigh()) {
+        PushCommand(FlushRegionCommand(addr, size));
+        return;
+    }
+    if (!Settings::IsGPULevelExtreme()) {
+        return;
+    }
+    if (system.Renderer().Rasterizer().MustFlushRegion(addr, size)) {
+        auto& gpu = system.GPU();
+        u64 fence = gpu.RequestFlush(addr, size);
+        PushCommand(GPUTickCommand());
+        while (fence > gpu.CurrentFlushRequestFence()) {
+        }
+    }
 }
 
 void ThreadManager::InvalidateRegion(VAddr addr, u64 size) {
-    system.Renderer().Rasterizer().InvalidateRegion(addr, size);
+    system.Renderer().Rasterizer().OnCPUWrite(addr, size);
 }
 
 void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
-    InvalidateRegion(addr, size);
+    system.Renderer().Rasterizer().OnCPUWrite(addr, size);
 }
 
 void ThreadManager::WaitIdle() const {
@@ -95,6 +114,10 @@ void ThreadManager::WaitIdle() const {
     }
 }
 
+void ThreadManager::OnCommandListEnd() {
+    PushCommand(OnCommandListEndCommand());
+}
+
 u64 ThreadManager::PushCommand(CommandData&& command_data) {
     const u64 fence{++state.last_fence};
     state.queue.Push(CommandDataContainer(std::move(command_data), fence));
diff --git a/src/video_core/gpu_thread.h b/src/video_core/gpu_thread.h
index cd74ad330..5a28335d6 100644
--- a/src/video_core/gpu_thread.h
+++ b/src/video_core/gpu_thread.h
@@ -70,9 +70,16 @@ struct FlushAndInvalidateRegionCommand final {
     u64 size;
 };
 
+/// Command called within the gpu, to schedule actions after a command list end
+struct OnCommandListEndCommand final {};
+
+/// Command to make the gpu look into pending requests
+struct GPUTickCommand final {};
+
 using CommandData =
     std::variant<EndProcessingCommand, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
-                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand>;
+                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand, OnCommandListEndCommand,
+                 GPUTickCommand>;
 
 struct CommandDataContainer {
     CommandDataContainer() = default;
@@ -122,6 +129,8 @@ public:
     // Wait until the gpu thread is idle.
     void WaitIdle() const;
 
+    void OnCommandListEnd();
+
 private:
     /// Pushes a command to be executed by the GPU thread
     u64 PushCommand(CommandData&& command_data);
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index fd49bc2a9..dbee9f634 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -51,11 +51,8 @@ GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, u64 size) {
     const GPUVAddr gpu_addr{FindFreeRegion(address_space_base, aligned_size)};
 
     MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr);
-    ASSERT(system.CurrentProcess()
-               ->PageTable()
-               .SetMemoryAttribute(cpu_addr, size, Kernel::Memory::MemoryAttribute::DeviceShared,
-                                   Kernel::Memory::MemoryAttribute::DeviceShared)
-               .IsSuccess());
+    ASSERT(
+        system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess());
 
     return gpu_addr;
 }
@@ -66,11 +63,8 @@ GPUVAddr MemoryManager::MapBufferEx(VAddr cpu_addr, GPUVAddr gpu_addr, u64 size)
     const u64 aligned_size{Common::AlignUp(size, page_size)};
 
     MapBackingMemory(gpu_addr, system.Memory().GetPointer(cpu_addr), aligned_size, cpu_addr);
-    ASSERT(system.CurrentProcess()
-               ->PageTable()
-               .SetMemoryAttribute(cpu_addr, size, Kernel::Memory::MemoryAttribute::DeviceShared,
-                                   Kernel::Memory::MemoryAttribute::DeviceShared)
-               .IsSuccess());
+    ASSERT(
+        system.CurrentProcess()->PageTable().LockForDeviceAddressSpace(cpu_addr, size).IsSuccess());
     return gpu_addr;
 }
 
@@ -87,9 +81,7 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) {
     UnmapRange(gpu_addr, aligned_size);
     ASSERT(system.CurrentProcess()
                ->PageTable()
-               .SetMemoryAttribute(cpu_addr.value(), size,
-                                   Kernel::Memory::MemoryAttribute::DeviceShared,
-                                   Kernel::Memory::MemoryAttribute::None)
+               .UnlockForDeviceAddressSpace(cpu_addr.value(), size)
                .IsSuccess());
 
     return gpu_addr;
diff --git a/src/video_core/query_cache.h b/src/video_core/query_cache.h
index 5ea2b01f2..2f75f8801 100644
--- a/src/video_core/query_cache.h
+++ b/src/video_core/query_cache.h
@@ -12,10 +12,12 @@
 #include <mutex>
 #include <optional>
 #include <unordered_map>
+#include <unordered_set>
 #include <vector>
 
 #include "common/assert.h"
 #include "core/core.h"
+#include "core/settings.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
@@ -130,6 +132,9 @@ public:
         }
 
         query->BindCounter(Stream(type).Current(), timestamp);
+        if (Settings::values.use_asynchronous_gpu_emulation) {
+            AsyncFlushQuery(cpu_addr);
+        }
     }
 
     /// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
@@ -170,6 +175,37 @@ public:
         return streams[static_cast<std::size_t>(type)];
     }
 
+    void CommitAsyncFlushes() {
+        committed_flushes.push_back(uncommitted_flushes);
+        uncommitted_flushes.reset();
+    }
+
+    bool HasUncommittedFlushes() const {
+        return uncommitted_flushes != nullptr;
+    }
+
+    bool ShouldWaitAsyncFlushes() const {
+        if (committed_flushes.empty()) {
+            return false;
+        }
+        return committed_flushes.front() != nullptr;
+    }
+
+    void PopAsyncFlushes() {
+        if (committed_flushes.empty()) {
+            return;
+        }
+        auto& flush_list = committed_flushes.front();
+        if (!flush_list) {
+            committed_flushes.pop_front();
+            return;
+        }
+        for (VAddr query_address : *flush_list) {
+            FlushAndRemoveRegion(query_address, 4);
+        }
+        committed_flushes.pop_front();
+    }
+
 protected:
     std::array<QueryPool, VideoCore::NumQueryTypes> query_pools;
 
@@ -224,6 +260,13 @@ private:
         return found != std::end(contents) ? &*found : nullptr;
     }
 
+    void AsyncFlushQuery(VAddr addr) {
+        if (!uncommitted_flushes) {
+            uncommitted_flushes = std::make_shared<std::unordered_set<VAddr>>();
+        }
+        uncommitted_flushes->insert(addr);
+    }
+
     static constexpr std::uintptr_t PAGE_SIZE = 4096;
     static constexpr unsigned PAGE_SHIFT = 12;
 
@@ -235,6 +278,9 @@ private:
     std::unordered_map<u64, std::vector<CachedQuery>> cached_queries;
 
     std::array<CounterStream, VideoCore::NumQueryTypes> streams;
+
+    std::shared_ptr<std::unordered_set<VAddr>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::unordered_set<VAddr>>> committed_flushes;
 };
 
 template <class QueryCache, class HostCounter>
diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h
index 8ae5b9c4e..603f61952 100644
--- a/src/video_core/rasterizer_interface.h
+++ b/src/video_core/rasterizer_interface.h
@@ -49,15 +49,33 @@ public:
     /// Records a GPU query and caches it
     virtual void Query(GPUVAddr gpu_addr, QueryType type, std::optional<u64> timestamp) = 0;
 
+    /// Signal a GPU based semaphore as a fence
+    virtual void SignalSemaphore(GPUVAddr addr, u32 value) = 0;
+
+    /// Signal a GPU based syncpoint as a fence
+    virtual void SignalSyncPoint(u32 value) = 0;
+
+    /// Release all pending fences.
+    virtual void ReleaseFences() = 0;
+
     /// Notify rasterizer that all caches should be flushed to Switch memory
     virtual void FlushAll() = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     virtual void FlushRegion(VAddr addr, u64 size) = 0;
 
+    /// Check if the the specified memory area requires flushing to CPU Memory.
+    virtual bool MustFlushRegion(VAddr addr, u64 size) = 0;
+
     /// Notify rasterizer that any caches of the specified region should be invalidated
     virtual void InvalidateRegion(VAddr addr, u64 size) = 0;
 
+    /// Notify rasterizer that any caches of the specified region are desync with guest
+    virtual void OnCPUWrite(VAddr addr, u64 size) = 0;
+
+    /// Sync memory between guest and host.
+    virtual void SyncGuestHost() = 0;
+
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     /// and invalidated
     virtual void FlushAndInvalidateRegion(VAddr addr, u64 size) = 0;
diff --git a/src/video_core/renderer_opengl/gl_buffer_cache.cpp b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
index cb5792407..4efce0de7 100644
--- a/src/video_core/renderer_opengl/gl_buffer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_buffer_cache.cpp
@@ -52,7 +52,7 @@ Buffer OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
 }
 
 void OGLBufferCache::WriteBarrier() {
-    glMemoryBarrier(GL_ALL_BARRIER_BITS);
+    glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT);
 }
 
 GLuint OGLBufferCache::ToHandle(const Buffer& buffer) {
@@ -72,6 +72,7 @@ void OGLBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, s
 void OGLBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                                        u8* data) {
     MICROPROFILE_SCOPE(OpenGL_Buffer_Download);
+    glMemoryBarrier(GL_BUFFER_UPDATE_BARRIER_BIT);
     glGetNamedBufferSubData(buffer->GetHandle(), static_cast<GLintptr>(offset),
                             static_cast<GLsizeiptr>(size), data);
 }
diff --git a/src/video_core/renderer_opengl/gl_fence_manager.cpp b/src/video_core/renderer_opengl/gl_fence_manager.cpp
new file mode 100644
index 000000000..99ddcb3f8
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_fence_manager.cpp
@@ -0,0 +1,72 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+
+#include "video_core/renderer_opengl/gl_fence_manager.h"
+
+namespace OpenGL {
+
+GLInnerFence::GLInnerFence(u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(payload, is_stubbed), sync_object{} {}
+
+GLInnerFence::GLInnerFence(GPUVAddr address, u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(address, payload, is_stubbed), sync_object{} {}
+
+GLInnerFence::~GLInnerFence() = default;
+
+void GLInnerFence::Queue() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(sync_object.handle == 0);
+    sync_object.Create();
+}
+
+bool GLInnerFence::IsSignaled() const {
+    if (is_stubbed) {
+        return true;
+    }
+    ASSERT(sync_object.handle != 0);
+    GLsizei length;
+    GLint sync_status;
+    glGetSynciv(sync_object.handle, GL_SYNC_STATUS, sizeof(GLint), &length, &sync_status);
+    return sync_status == GL_SIGNALED;
+}
+
+void GLInnerFence::Wait() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(sync_object.handle != 0);
+    glClientWaitSync(sync_object.handle, 0, GL_TIMEOUT_IGNORED);
+}
+
+FenceManagerOpenGL::FenceManagerOpenGL(Core::System& system,
+                                       VideoCore::RasterizerInterface& rasterizer,
+                                       TextureCacheOpenGL& texture_cache,
+                                       OGLBufferCache& buffer_cache, QueryCache& query_cache)
+    : GenericFenceManager(system, rasterizer, texture_cache, buffer_cache, query_cache) {}
+
+Fence FenceManagerOpenGL::CreateFence(u32 value, bool is_stubbed) {
+    return std::make_shared<GLInnerFence>(value, is_stubbed);
+}
+
+Fence FenceManagerOpenGL::CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) {
+    return std::make_shared<GLInnerFence>(addr, value, is_stubbed);
+}
+
+void FenceManagerOpenGL::QueueFence(Fence& fence) {
+    fence->Queue();
+}
+
+bool FenceManagerOpenGL::IsFenceSignaled(Fence& fence) const {
+    return fence->IsSignaled();
+}
+
+void FenceManagerOpenGL::WaitFence(Fence& fence) {
+    fence->Wait();
+}
+
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_fence_manager.h b/src/video_core/renderer_opengl/gl_fence_manager.h
new file mode 100644
index 000000000..c917b3343
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_fence_manager.h
@@ -0,0 +1,53 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <glad/glad.h>
+
+#include "common/common_types.h"
+#include "video_core/fence_manager.h"
+#include "video_core/renderer_opengl/gl_buffer_cache.h"
+#include "video_core/renderer_opengl/gl_query_cache.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/renderer_opengl/gl_texture_cache.h"
+
+namespace OpenGL {
+
+class GLInnerFence : public VideoCommon::FenceBase {
+public:
+    GLInnerFence(u32 payload, bool is_stubbed);
+    GLInnerFence(GPUVAddr address, u32 payload, bool is_stubbed);
+    ~GLInnerFence();
+
+    void Queue();
+
+    bool IsSignaled() const;
+
+    void Wait();
+
+private:
+    OGLSync sync_object;
+};
+
+using Fence = std::shared_ptr<GLInnerFence>;
+using GenericFenceManager =
+    VideoCommon::FenceManager<Fence, TextureCacheOpenGL, OGLBufferCache, QueryCache>;
+
+class FenceManagerOpenGL final : public GenericFenceManager {
+public:
+    FenceManagerOpenGL(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                       TextureCacheOpenGL& texture_cache, OGLBufferCache& buffer_cache,
+                       QueryCache& query_cache);
+
+protected:
+    Fence CreateFence(u32 value, bool is_stubbed) override;
+    Fence CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) override;
+    void QueueFence(Fence& fence) override;
+    bool IsFenceSignaled(Fence& fence) const override;
+    void WaitFence(Fence& fence) override;
+};
+
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index 175374f0d..6fe155bcc 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -99,9 +99,10 @@ RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWind
                                    ScreenInfo& info, GLShader::ProgramManager& program_manager,
                                    StateTracker& state_tracker)
     : RasterizerAccelerated{system.Memory()}, texture_cache{system, *this, device, state_tracker},
-      shader_cache{*this, system, emu_window, device}, query_cache{system, *this}, system{system},
-      screen_info{info}, program_manager{program_manager}, state_tracker{state_tracker},
-      buffer_cache{*this, system, device, STREAM_BUFFER_SIZE} {
+      shader_cache{*this, system, emu_window, device}, query_cache{system, *this},
+      buffer_cache{*this, system, device, STREAM_BUFFER_SIZE},
+      fence_manager{system, *this, texture_cache, buffer_cache, query_cache}, system{system},
+      screen_info{info}, program_manager{program_manager}, state_tracker{state_tracker} {
     CheckExtensions();
 }
 
@@ -185,8 +186,12 @@ void RasterizerOpenGL::SetupVertexBuffer() {
         const GPUVAddr start = vertex_array.StartAddress();
         const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
 
-        ASSERT(end > start);
-        const u64 size = end - start + 1;
+        ASSERT(end >= start);
+        const u64 size = end - start;
+        if (size == 0) {
+            glBindVertexBuffer(static_cast<GLuint>(index), 0, 0, vertex_array.stride);
+            continue;
+        }
         const auto [vertex_buffer, vertex_buffer_offset] = buffer_cache.UploadMemory(start, size);
         glBindVertexBuffer(static_cast<GLuint>(index), vertex_buffer, vertex_buffer_offset,
                            vertex_array.stride);
@@ -310,8 +315,8 @@ std::size_t RasterizerOpenGL::CalculateVertexArraysSize() const {
         const GPUVAddr start = regs.vertex_array[index].StartAddress();
         const GPUVAddr end = regs.vertex_array_limit[index].LimitAddress();
 
-        ASSERT(end > start);
-        size += end - start + 1;
+        size += end - start;
+        ASSERT(end >= start);
     }
 
     return size;
@@ -599,6 +604,8 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     EndTransformFeedback();
 
     ++num_queued_commands;
+
+    system.GPU().TickWork();
 }
 
 void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
@@ -649,6 +656,13 @@ void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size) {
     query_cache.FlushRegion(addr, size);
 }
 
+bool RasterizerOpenGL::MustFlushRegion(VAddr addr, u64 size) {
+    if (!Settings::IsGPULevelHigh()) {
+        return buffer_cache.MustFlushRegion(addr, size);
+    }
+    return texture_cache.MustFlushRegion(addr, size) || buffer_cache.MustFlushRegion(addr, size);
+}
+
 void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     if (addr == 0 || size == 0) {
@@ -660,8 +674,52 @@ void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size) {
     query_cache.InvalidateRegion(addr, size);
 }
 
+void RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    if (addr == 0 || size == 0) {
+        return;
+    }
+    texture_cache.OnCPUWrite(addr, size);
+    shader_cache.InvalidateRegion(addr, size);
+    buffer_cache.OnCPUWrite(addr, size);
+    query_cache.InvalidateRegion(addr, size);
+}
+
+void RasterizerOpenGL::SyncGuestHost() {
+    MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    texture_cache.SyncGuestHost();
+    buffer_cache.SyncGuestHost();
+}
+
+void RasterizerOpenGL::SignalSemaphore(GPUVAddr addr, u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        auto& memory_manager{gpu.MemoryManager()};
+        memory_manager.Write<u32>(addr, value);
+        return;
+    }
+    fence_manager.SignalSemaphore(addr, value);
+}
+
+void RasterizerOpenGL::SignalSyncPoint(u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        gpu.IncrementSyncPoint(value);
+        return;
+    }
+    fence_manager.SignalSyncPoint(value);
+}
+
+void RasterizerOpenGL::ReleaseFences() {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        return;
+    }
+    fence_manager.WaitPendingFences();
+}
+
 void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
-    if (Settings::values.use_accurate_gpu_emulation) {
+    if (Settings::IsGPULevelExtreme()) {
         FlushRegion(addr, size);
     }
     InvalidateRegion(addr, size);
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index caea174d2..ebd2173eb 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -23,6 +23,7 @@
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_device.h"
+#include "video_core/renderer_opengl/gl_fence_manager.h"
 #include "video_core/renderer_opengl/gl_framebuffer_cache.h"
 #include "video_core/renderer_opengl/gl_query_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
@@ -66,7 +67,13 @@ public:
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
     void FlushAll() override;
     void FlushRegion(VAddr addr, u64 size) override;
+    bool MustFlushRegion(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size) override;
+    void OnCPUWrite(VAddr addr, u64 size) override;
+    void SyncGuestHost() override;
+    void SignalSemaphore(GPUVAddr addr, u32 value) override;
+    void SignalSyncPoint(u32 value) override;
+    void ReleaseFences() override;
     void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void FlushCommands() override;
     void TickFrame() override;
@@ -222,6 +229,8 @@ private:
     SamplerCacheOpenGL sampler_cache;
     FramebufferCacheOpenGL framebuffer_cache;
     QueryCache query_cache;
+    OGLBufferCache buffer_cache;
+    FenceManagerOpenGL fence_manager;
 
     Core::System& system;
     ScreenInfo& screen_info;
@@ -229,7 +238,6 @@ private:
     StateTracker& state_tracker;
 
     static constexpr std::size_t STREAM_BUFFER_SIZE = 128 * 1024 * 1024;
-    OGLBufferCache buffer_cache;
 
     GLint vertex_binding = 0;
 
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.cpp b/src/video_core/renderer_opengl/gl_shader_cache.cpp
index 6d2ff20f9..f63156b8d 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_cache.cpp
@@ -448,7 +448,7 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
 
     // Look up shader in the cache based on address
     const auto cpu_addr{memory_manager.GpuToCpuAddress(address)};
-    Shader shader{cpu_addr ? TryGet(*cpu_addr) : nullptr};
+    Shader shader{cpu_addr ? TryGet(*cpu_addr) : null_shader};
     if (shader) {
         return last_shaders[static_cast<std::size_t>(program)] = shader;
     }
@@ -477,7 +477,12 @@ Shader ShaderCacheOpenGL::GetStageProgram(Maxwell::ShaderProgram program) {
         const std::size_t size_in_bytes = code.size() * sizeof(u64);
         shader = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
     }
-    Register(shader);
+
+    if (cpu_addr) {
+        Register(shader);
+    } else {
+        null_shader = shader;
+    }
 
     return last_shaders[static_cast<std::size_t>(program)] = shader;
 }
@@ -486,7 +491,7 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
     auto& memory_manager{system.GPU().MemoryManager()};
     const auto cpu_addr{memory_manager.GpuToCpuAddress(code_addr)};
 
-    auto kernel = cpu_addr ? TryGet(*cpu_addr) : nullptr;
+    auto kernel = cpu_addr ? TryGet(*cpu_addr) : null_kernel;
     if (kernel) {
         return kernel;
     }
@@ -507,7 +512,11 @@ Shader ShaderCacheOpenGL::GetComputeKernel(GPUVAddr code_addr) {
         kernel = CachedShader::CreateFromCache(params, found->second, size_in_bytes);
     }
 
-    Register(kernel);
+    if (cpu_addr) {
+        Register(kernel);
+    } else {
+        null_kernel = kernel;
+    }
     return kernel;
 }
 
diff --git a/src/video_core/renderer_opengl/gl_shader_cache.h b/src/video_core/renderer_opengl/gl_shader_cache.h
index c836df5bd..91690b470 100644
--- a/src/video_core/renderer_opengl/gl_shader_cache.h
+++ b/src/video_core/renderer_opengl/gl_shader_cache.h
@@ -125,6 +125,9 @@ private:
     ShaderDiskCacheOpenGL disk_cache;
     std::unordered_map<u64, PrecompiledShader> runtime_cache;
 
+    Shader null_shader{};
+    Shader null_kernel{};
+
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 };
 
diff --git a/src/video_core/renderer_vulkan/fixed_pipeline_state.h b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
index 9fe6bdbf9..9a950f4de 100644
--- a/src/video_core/renderer_vulkan/fixed_pipeline_state.h
+++ b/src/video_core/renderer_vulkan/fixed_pipeline_state.h
@@ -129,7 +129,7 @@ struct alignas(32) FixedPipelineState {
             auto& binding = bindings[index];
             binding.raw = 0;
             binding.enabled.Assign(enabled ? 1 : 0);
-            binding.stride.Assign(stride);
+            binding.stride.Assign(static_cast<u16>(stride));
             binding_divisors[index] = divisor;
         }
 
diff --git a/src/video_core/renderer_vulkan/nsight_aftermath_tracker.cpp b/src/video_core/renderer_vulkan/nsight_aftermath_tracker.cpp
new file mode 100644
index 000000000..435c8c1b8
--- /dev/null
+++ b/src/video_core/renderer_vulkan/nsight_aftermath_tracker.cpp
@@ -0,0 +1,220 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#ifdef HAS_NSIGHT_AFTERMATH
+
+#include <mutex>
+#include <string>
+#include <string_view>
+#include <utility>
+#include <vector>
+
+#include <fmt/format.h>
+
+#define VK_NO_PROTOTYPES
+#include <vulkan/vulkan.h>
+
+#include <GFSDK_Aftermath.h>
+#include <GFSDK_Aftermath_Defines.h>
+#include <GFSDK_Aftermath_GpuCrashDump.h>
+#include <GFSDK_Aftermath_GpuCrashDumpDecoding.h>
+
+#include "common/common_paths.h"
+#include "common/common_types.h"
+#include "common/file_util.h"
+#include "common/logging/log.h"
+#include "common/scope_exit.h"
+
+#include "video_core/renderer_vulkan/nsight_aftermath_tracker.h"
+
+namespace Vulkan {
+
+static constexpr char AFTERMATH_LIB_NAME[] = "GFSDK_Aftermath_Lib.x64.dll";
+
+NsightAftermathTracker::NsightAftermathTracker() = default;
+
+NsightAftermathTracker::~NsightAftermathTracker() {
+    if (initialized) {
+        (void)GFSDK_Aftermath_DisableGpuCrashDumps();
+    }
+}
+
+bool NsightAftermathTracker::Initialize() {
+    if (!dl.Open(AFTERMATH_LIB_NAME)) {
+        LOG_ERROR(Render_Vulkan, "Failed to load Nsight Aftermath DLL");
+        return false;
+    }
+
+    if (!dl.GetSymbol("GFSDK_Aftermath_DisableGpuCrashDumps",
+                      &GFSDK_Aftermath_DisableGpuCrashDumps) ||
+        !dl.GetSymbol("GFSDK_Aftermath_EnableGpuCrashDumps",
+                      &GFSDK_Aftermath_EnableGpuCrashDumps) ||
+        !dl.GetSymbol("GFSDK_Aftermath_GetShaderDebugInfoIdentifier",
+                      &GFSDK_Aftermath_GetShaderDebugInfoIdentifier) ||
+        !dl.GetSymbol("GFSDK_Aftermath_GetShaderHashSpirv", &GFSDK_Aftermath_GetShaderHashSpirv) ||
+        !dl.GetSymbol("GFSDK_Aftermath_GpuCrashDump_CreateDecoder",
+                      &GFSDK_Aftermath_GpuCrashDump_CreateDecoder) ||
+        !dl.GetSymbol("GFSDK_Aftermath_GpuCrashDump_DestroyDecoder",
+                      &GFSDK_Aftermath_GpuCrashDump_DestroyDecoder) ||
+        !dl.GetSymbol("GFSDK_Aftermath_GpuCrashDump_GenerateJSON",
+                      &GFSDK_Aftermath_GpuCrashDump_GenerateJSON) ||
+        !dl.GetSymbol("GFSDK_Aftermath_GpuCrashDump_GetJSON",
+                      &GFSDK_Aftermath_GpuCrashDump_GetJSON)) {
+        LOG_ERROR(Render_Vulkan, "Failed to load Nsight Aftermath function pointers");
+        return false;
+    }
+
+    dump_dir = FileUtil::GetUserPath(FileUtil::UserPath::LogDir) + "gpucrash";
+
+    (void)FileUtil::DeleteDirRecursively(dump_dir);
+    if (!FileUtil::CreateDir(dump_dir)) {
+        LOG_ERROR(Render_Vulkan, "Failed to create Nsight Aftermath dump directory");
+        return false;
+    }
+
+    if (!GFSDK_Aftermath_SUCCEED(GFSDK_Aftermath_EnableGpuCrashDumps(
+            GFSDK_Aftermath_Version_API, GFSDK_Aftermath_GpuCrashDumpWatchedApiFlags_Vulkan,
+            GFSDK_Aftermath_GpuCrashDumpFeatureFlags_Default, GpuCrashDumpCallback,
+            ShaderDebugInfoCallback, CrashDumpDescriptionCallback, this))) {
+        LOG_ERROR(Render_Vulkan, "GFSDK_Aftermath_EnableGpuCrashDumps failed");
+        return false;
+    }
+
+    LOG_INFO(Render_Vulkan, "Nsight Aftermath dump directory is \"{}\"", dump_dir);
+
+    initialized = true;
+    return true;
+}
+
+void NsightAftermathTracker::SaveShader(const std::vector<u32>& spirv) const {
+    if (!initialized) {
+        return;
+    }
+
+    std::vector<u32> spirv_copy = spirv;
+    GFSDK_Aftermath_SpirvCode shader;
+    shader.pData = spirv_copy.data();
+    shader.size = static_cast<u32>(spirv_copy.size() * 4);
+
+    std::scoped_lock lock{mutex};
+
+    GFSDK_Aftermath_ShaderHash hash;
+    if (!GFSDK_Aftermath_SUCCEED(
+            GFSDK_Aftermath_GetShaderHashSpirv(GFSDK_Aftermath_Version_API, &shader, &hash))) {
+        LOG_ERROR(Render_Vulkan, "Failed to hash SPIR-V module");
+        return;
+    }
+
+    FileUtil::IOFile file(fmt::format("{}/source_{:016x}.spv", dump_dir, hash.hash), "wb");
+    if (!file.IsOpen()) {
+        LOG_ERROR(Render_Vulkan, "Failed to dump SPIR-V module with hash={:016x}", hash.hash);
+        return;
+    }
+    if (file.WriteArray(spirv.data(), spirv.size()) != spirv.size()) {
+        LOG_ERROR(Render_Vulkan, "Failed to write SPIR-V module with hash={:016x}", hash.hash);
+        return;
+    }
+}
+
+void NsightAftermathTracker::OnGpuCrashDumpCallback(const void* gpu_crash_dump,
+                                                    u32 gpu_crash_dump_size) {
+    std::scoped_lock lock{mutex};
+
+    LOG_CRITICAL(Render_Vulkan, "called");
+
+    GFSDK_Aftermath_GpuCrashDump_Decoder decoder;
+    if (!GFSDK_Aftermath_SUCCEED(GFSDK_Aftermath_GpuCrashDump_CreateDecoder(
+            GFSDK_Aftermath_Version_API, gpu_crash_dump, gpu_crash_dump_size, &decoder))) {
+        LOG_ERROR(Render_Vulkan, "Failed to create decoder");
+        return;
+    }
+    SCOPE_EXIT({ GFSDK_Aftermath_GpuCrashDump_DestroyDecoder(decoder); });
+
+    u32 json_size = 0;
+    if (!GFSDK_Aftermath_SUCCEED(GFSDK_Aftermath_GpuCrashDump_GenerateJSON(
+            decoder, GFSDK_Aftermath_GpuCrashDumpDecoderFlags_ALL_INFO,
+            GFSDK_Aftermath_GpuCrashDumpFormatterFlags_NONE, nullptr, nullptr, nullptr, nullptr,
+            this, &json_size))) {
+        LOG_ERROR(Render_Vulkan, "Failed to generate JSON");
+        return;
+    }
+    std::vector<char> json(json_size);
+    if (!GFSDK_Aftermath_SUCCEED(
+            GFSDK_Aftermath_GpuCrashDump_GetJSON(decoder, json_size, json.data()))) {
+        LOG_ERROR(Render_Vulkan, "Failed to query JSON");
+        return;
+    }
+
+    const std::string base_name = [this] {
+        const int id = dump_id++;
+        if (id == 0) {
+            return fmt::format("{}/crash.nv-gpudmp", dump_dir);
+        } else {
+            return fmt::format("{}/crash_{}.nv-gpudmp", dump_dir, id);
+        }
+    }();
+
+    std::string_view dump_view(static_cast<const char*>(gpu_crash_dump), gpu_crash_dump_size);
+    if (FileUtil::WriteStringToFile(false, base_name, dump_view) != gpu_crash_dump_size) {
+        LOG_ERROR(Render_Vulkan, "Failed to write dump file");
+        return;
+    }
+    const std::string_view json_view(json.data(), json.size());
+    if (FileUtil::WriteStringToFile(true, base_name + ".json", json_view) != json.size()) {
+        LOG_ERROR(Render_Vulkan, "Failed to write JSON");
+        return;
+    }
+}
+
+void NsightAftermathTracker::OnShaderDebugInfoCallback(const void* shader_debug_info,
+                                                       u32 shader_debug_info_size) {
+    std::scoped_lock lock{mutex};
+
+    GFSDK_Aftermath_ShaderDebugInfoIdentifier identifier;
+    if (!GFSDK_Aftermath_SUCCEED(GFSDK_Aftermath_GetShaderDebugInfoIdentifier(
+            GFSDK_Aftermath_Version_API, shader_debug_info, shader_debug_info_size, &identifier))) {
+        LOG_ERROR(Render_Vulkan, "GFSDK_Aftermath_GetShaderDebugInfoIdentifier failed");
+        return;
+    }
+
+    const std::string path =
+        fmt::format("{}/shader_{:016x}{:016x}.nvdbg", dump_dir, identifier.id[0], identifier.id[1]);
+    FileUtil::IOFile file(path, "wb");
+    if (!file.IsOpen()) {
+        LOG_ERROR(Render_Vulkan, "Failed to create file {}", path);
+        return;
+    }
+    if (file.WriteBytes(static_cast<const u8*>(shader_debug_info), shader_debug_info_size) !=
+        shader_debug_info_size) {
+        LOG_ERROR(Render_Vulkan, "Failed to write file {}", path);
+        return;
+    }
+}
+
+void NsightAftermathTracker::OnCrashDumpDescriptionCallback(
+    PFN_GFSDK_Aftermath_AddGpuCrashDumpDescription add_description) {
+    add_description(GFSDK_Aftermath_GpuCrashDumpDescriptionKey_ApplicationName, "yuzu");
+}
+
+void NsightAftermathTracker::GpuCrashDumpCallback(const void* gpu_crash_dump,
+                                                  u32 gpu_crash_dump_size, void* user_data) {
+    static_cast<NsightAftermathTracker*>(user_data)->OnGpuCrashDumpCallback(gpu_crash_dump,
+                                                                            gpu_crash_dump_size);
+}
+
+void NsightAftermathTracker::ShaderDebugInfoCallback(const void* shader_debug_info,
+                                                     u32 shader_debug_info_size, void* user_data) {
+    static_cast<NsightAftermathTracker*>(user_data)->OnShaderDebugInfoCallback(
+        shader_debug_info, shader_debug_info_size);
+}
+
+void NsightAftermathTracker::CrashDumpDescriptionCallback(
+    PFN_GFSDK_Aftermath_AddGpuCrashDumpDescription add_description, void* user_data) {
+    static_cast<NsightAftermathTracker*>(user_data)->OnCrashDumpDescriptionCallback(
+        add_description);
+}
+
+} // namespace Vulkan
+
+#endif // HAS_NSIGHT_AFTERMATH
diff --git a/src/video_core/renderer_vulkan/nsight_aftermath_tracker.h b/src/video_core/renderer_vulkan/nsight_aftermath_tracker.h
new file mode 100644
index 000000000..afe7ae99e
--- /dev/null
+++ b/src/video_core/renderer_vulkan/nsight_aftermath_tracker.h
@@ -0,0 +1,87 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <mutex>
+#include <string>
+#include <vector>
+
+#define VK_NO_PROTOTYPES
+#include <vulkan/vulkan.h>
+
+#ifdef HAS_NSIGHT_AFTERMATH
+#include <GFSDK_Aftermath_Defines.h>
+#include <GFSDK_Aftermath_GpuCrashDump.h>
+#include <GFSDK_Aftermath_GpuCrashDumpDecoding.h>
+#endif
+
+#include "common/common_types.h"
+#include "common/dynamic_library.h"
+
+namespace Vulkan {
+
+class NsightAftermathTracker {
+public:
+    NsightAftermathTracker();
+    ~NsightAftermathTracker();
+
+    NsightAftermathTracker(const NsightAftermathTracker&) = delete;
+    NsightAftermathTracker& operator=(const NsightAftermathTracker&) = delete;
+
+    // Delete move semantics because Aftermath initialization uses a pointer to this.
+    NsightAftermathTracker(NsightAftermathTracker&&) = delete;
+    NsightAftermathTracker& operator=(NsightAftermathTracker&&) = delete;
+
+    bool Initialize();
+
+    void SaveShader(const std::vector<u32>& spirv) const;
+
+private:
+#ifdef HAS_NSIGHT_AFTERMATH
+    static void GpuCrashDumpCallback(const void* gpu_crash_dump, u32 gpu_crash_dump_size,
+                                     void* user_data);
+
+    static void ShaderDebugInfoCallback(const void* shader_debug_info, u32 shader_debug_info_size,
+                                        void* user_data);
+
+    static void CrashDumpDescriptionCallback(
+        PFN_GFSDK_Aftermath_AddGpuCrashDumpDescription add_description, void* user_data);
+
+    void OnGpuCrashDumpCallback(const void* gpu_crash_dump, u32 gpu_crash_dump_size);
+
+    void OnShaderDebugInfoCallback(const void* shader_debug_info, u32 shader_debug_info_size);
+
+    void OnCrashDumpDescriptionCallback(
+        PFN_GFSDK_Aftermath_AddGpuCrashDumpDescription add_description);
+
+    mutable std::mutex mutex;
+
+    std::string dump_dir;
+    int dump_id = 0;
+
+    bool initialized = false;
+
+    Common::DynamicLibrary dl;
+    PFN_GFSDK_Aftermath_DisableGpuCrashDumps GFSDK_Aftermath_DisableGpuCrashDumps;
+    PFN_GFSDK_Aftermath_EnableGpuCrashDumps GFSDK_Aftermath_EnableGpuCrashDumps;
+    PFN_GFSDK_Aftermath_GetShaderDebugInfoIdentifier GFSDK_Aftermath_GetShaderDebugInfoIdentifier;
+    PFN_GFSDK_Aftermath_GetShaderHashSpirv GFSDK_Aftermath_GetShaderHashSpirv;
+    PFN_GFSDK_Aftermath_GpuCrashDump_CreateDecoder GFSDK_Aftermath_GpuCrashDump_CreateDecoder;
+    PFN_GFSDK_Aftermath_GpuCrashDump_DestroyDecoder GFSDK_Aftermath_GpuCrashDump_DestroyDecoder;
+    PFN_GFSDK_Aftermath_GpuCrashDump_GenerateJSON GFSDK_Aftermath_GpuCrashDump_GenerateJSON;
+    PFN_GFSDK_Aftermath_GpuCrashDump_GetJSON GFSDK_Aftermath_GpuCrashDump_GetJSON;
+#endif
+};
+
+#ifndef HAS_NSIGHT_AFTERMATH
+inline NsightAftermathTracker::NsightAftermathTracker() = default;
+inline NsightAftermathTracker::~NsightAftermathTracker() = default;
+inline bool NsightAftermathTracker::Initialize() {
+    return false;
+}
+inline void NsightAftermathTracker::SaveShader(const std::vector<u32>&) const {}
+#endif
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
index 23beafa4f..52566bb79 100644
--- a/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_compute_pipeline.cpp
@@ -105,6 +105,8 @@ vk::DescriptorUpdateTemplateKHR VKComputePipeline::CreateDescriptorUpdateTemplat
 }
 
 vk::ShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const {
+    device.SaveShader(code);
+
     VkShaderModuleCreateInfo ci;
     ci.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
     ci.pNext = nullptr;
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index 52d29e49d..e90c76492 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -9,6 +9,7 @@
 #include <string_view>
 #include <thread>
 #include <unordered_set>
+#include <utility>
 #include <vector>
 
 #include "common/assert.h"
@@ -167,6 +168,7 @@ bool VKDevice::Create() {
     VkPhysicalDeviceFeatures2 features2;
     features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2;
     features2.pNext = nullptr;
+    const void* first_next = &features2;
     void** next = &features2.pNext;
 
     auto& features = features2.features;
@@ -296,7 +298,19 @@ bool VKDevice::Create() {
         LOG_INFO(Render_Vulkan, "Device doesn't support depth range unrestricted");
     }
 
-    logical = vk::Device::Create(physical, queue_cis, extensions, features2, dld);
+    VkDeviceDiagnosticsConfigCreateInfoNV diagnostics_nv;
+    if (nv_device_diagnostics_config) {
+        nsight_aftermath_tracker.Initialize();
+
+        diagnostics_nv.sType = VK_STRUCTURE_TYPE_DEVICE_DIAGNOSTICS_CONFIG_CREATE_INFO_NV;
+        diagnostics_nv.pNext = &features2;
+        diagnostics_nv.flags = VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_SHADER_DEBUG_INFO_BIT_NV |
+                               VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_RESOURCE_TRACKING_BIT_NV |
+                               VK_DEVICE_DIAGNOSTICS_CONFIG_ENABLE_AUTOMATIC_CHECKPOINTS_BIT_NV;
+        first_next = &diagnostics_nv;
+    }
+
+    logical = vk::Device::Create(physical, queue_cis, extensions, first_next, dld);
     if (!logical) {
         LOG_ERROR(Render_Vulkan, "Failed to create logical device");
         return false;
@@ -344,17 +358,12 @@ VkFormat VKDevice::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFla
 void VKDevice::ReportLoss() const {
     LOG_CRITICAL(Render_Vulkan, "Device loss occured!");
 
-    // Wait some time to let the log flush
-    std::this_thread::sleep_for(std::chrono::seconds{1});
-
-    if (!nv_device_diagnostic_checkpoints) {
-        return;
-    }
+    // Wait for the log to flush and for Nsight Aftermath to dump the results
+    std::this_thread::sleep_for(std::chrono::seconds{3});
+}
 
-    [[maybe_unused]] const std::vector data = graphics_queue.GetCheckpointDataNV(dld);
-    // Catch here in debug builds (or with optimizations disabled) the last graphics pipeline to be
-    // executed. It can be done on a debugger by evaluating the expression:
-    // *(VKGraphicsPipeline*)data[0]
+void VKDevice::SaveShader(const std::vector<u32>& spirv) const {
+    nsight_aftermath_tracker.SaveShader(spirv);
 }
 
 bool VKDevice::IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const {
@@ -527,8 +536,8 @@ std::vector<const char*> VKDevice::LoadExtensions() {
         Test(extension, has_ext_transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME,
              false);
         if (Settings::values.renderer_debug) {
-            Test(extension, nv_device_diagnostic_checkpoints,
-                 VK_NV_DEVICE_DIAGNOSTIC_CHECKPOINTS_EXTENSION_NAME, true);
+            Test(extension, nv_device_diagnostics_config,
+                 VK_NV_DEVICE_DIAGNOSTICS_CONFIG_EXTENSION_NAME, true);
         }
     }
 
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index bac58d385..c8640762d 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -10,6 +10,7 @@
 #include <vector>
 
 #include "common/common_types.h"
+#include "video_core/renderer_vulkan/nsight_aftermath_tracker.h"
 #include "video_core/renderer_vulkan/wrapper.h"
 
 namespace Vulkan {
@@ -43,6 +44,9 @@ public:
     /// Reports a device loss.
     void ReportLoss() const;
 
+    /// Reports a shader to Nsight Aftermath.
+    void SaveShader(const std::vector<u32>& spirv) const;
+
     /// Returns the dispatch loader with direct function pointers of the device.
     const vk::DeviceDispatch& GetDispatchLoader() const {
         return dld;
@@ -168,11 +172,6 @@ public:
         return ext_transform_feedback;
     }
 
-    /// Returns true if the device supports VK_NV_device_diagnostic_checkpoints.
-    bool IsNvDeviceDiagnosticCheckpoints() const {
-        return nv_device_diagnostic_checkpoints;
-    }
-
     /// Returns the vendor name reported from Vulkan.
     std::string_view GetVendorName() const {
         return vendor_name;
@@ -228,7 +227,7 @@ private:
     bool ext_depth_range_unrestricted{};       ///< Support for VK_EXT_depth_range_unrestricted.
     bool ext_shader_viewport_index_layer{};    ///< Support for VK_EXT_shader_viewport_index_layer.
     bool ext_transform_feedback{};             ///< Support for VK_EXT_transform_feedback.
-    bool nv_device_diagnostic_checkpoints{};   ///< Support for VK_NV_device_diagnostic_checkpoints.
+    bool nv_device_diagnostics_config{};       ///< Support for VK_NV_device_diagnostics_config.
 
     // Telemetry parameters
     std::string vendor_name;                      ///< Device's driver name.
@@ -236,6 +235,9 @@ private:
 
     /// Format properties dictionary.
     std::unordered_map<VkFormat, VkFormatProperties> format_properties;
+
+    /// Nsight Aftermath GPU crash tracker
+    NsightAftermathTracker nsight_aftermath_tracker;
 };
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_fence_manager.cpp b/src/video_core/renderer_vulkan/vk_fence_manager.cpp
new file mode 100644
index 000000000..a02be5487
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_fence_manager.cpp
@@ -0,0 +1,101 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <memory>
+#include <thread>
+
+#include "video_core/renderer_vulkan/vk_buffer_cache.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_fence_manager.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_texture_cache.h"
+#include "video_core/renderer_vulkan/wrapper.h"
+
+namespace Vulkan {
+
+InnerFence::InnerFence(const VKDevice& device, VKScheduler& scheduler, u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(payload, is_stubbed), device{device}, scheduler{scheduler} {}
+
+InnerFence::InnerFence(const VKDevice& device, VKScheduler& scheduler, GPUVAddr address,
+                       u32 payload, bool is_stubbed)
+    : VideoCommon::FenceBase(address, payload, is_stubbed), device{device}, scheduler{scheduler} {}
+
+InnerFence::~InnerFence() = default;
+
+void InnerFence::Queue() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(!event);
+
+    event = device.GetLogical().CreateEvent();
+    ticks = scheduler.Ticks();
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([event = *event](vk::CommandBuffer cmdbuf) {
+        cmdbuf.SetEvent(event, VK_PIPELINE_STAGE_ALL_COMMANDS_BIT);
+    });
+}
+
+bool InnerFence::IsSignaled() const {
+    if (is_stubbed) {
+        return true;
+    }
+    ASSERT(event);
+    return IsEventSignalled();
+}
+
+void InnerFence::Wait() {
+    if (is_stubbed) {
+        return;
+    }
+    ASSERT(event);
+
+    if (ticks >= scheduler.Ticks()) {
+        scheduler.Flush();
+    }
+    while (!IsEventSignalled()) {
+        std::this_thread::yield();
+    }
+}
+
+bool InnerFence::IsEventSignalled() const {
+    switch (const VkResult result = event.GetStatus()) {
+    case VK_EVENT_SET:
+        return true;
+    case VK_EVENT_RESET:
+        return false;
+    default:
+        throw vk::Exception(result);
+    }
+}
+
+VKFenceManager::VKFenceManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                               const VKDevice& device, VKScheduler& scheduler,
+                               VKTextureCache& texture_cache, VKBufferCache& buffer_cache,
+                               VKQueryCache& query_cache)
+    : GenericFenceManager(system, rasterizer, texture_cache, buffer_cache, query_cache),
+      device{device}, scheduler{scheduler} {}
+
+Fence VKFenceManager::CreateFence(u32 value, bool is_stubbed) {
+    return std::make_shared<InnerFence>(device, scheduler, value, is_stubbed);
+}
+
+Fence VKFenceManager::CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) {
+    return std::make_shared<InnerFence>(device, scheduler, addr, value, is_stubbed);
+}
+
+void VKFenceManager::QueueFence(Fence& fence) {
+    fence->Queue();
+}
+
+bool VKFenceManager::IsFenceSignaled(Fence& fence) const {
+    return fence->IsSignaled();
+}
+
+void VKFenceManager::WaitFence(Fence& fence) {
+    fence->Wait();
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_fence_manager.h b/src/video_core/renderer_vulkan/vk_fence_manager.h
new file mode 100644
index 000000000..04d07fe6a
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_fence_manager.h
@@ -0,0 +1,74 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+
+#include "video_core/fence_manager.h"
+#include "video_core/renderer_vulkan/wrapper.h"
+
+namespace Core {
+class System;
+}
+
+namespace VideoCore {
+class RasterizerInterface;
+}
+
+namespace Vulkan {
+
+class VKBufferCache;
+class VKDevice;
+class VKQueryCache;
+class VKScheduler;
+class VKTextureCache;
+
+class InnerFence : public VideoCommon::FenceBase {
+public:
+    explicit InnerFence(const VKDevice& device, VKScheduler& scheduler, u32 payload,
+                        bool is_stubbed);
+    explicit InnerFence(const VKDevice& device, VKScheduler& scheduler, GPUVAddr address,
+                        u32 payload, bool is_stubbed);
+    ~InnerFence();
+
+    void Queue();
+
+    bool IsSignaled() const;
+
+    void Wait();
+
+private:
+    bool IsEventSignalled() const;
+
+    const VKDevice& device;
+    VKScheduler& scheduler;
+    vk::Event event;
+    u64 ticks = 0;
+};
+using Fence = std::shared_ptr<InnerFence>;
+
+using GenericFenceManager =
+    VideoCommon::FenceManager<Fence, VKTextureCache, VKBufferCache, VKQueryCache>;
+
+class VKFenceManager final : public GenericFenceManager {
+public:
+    explicit VKFenceManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                            const VKDevice& device, VKScheduler& scheduler,
+                            VKTextureCache& texture_cache, VKBufferCache& buffer_cache,
+                            VKQueryCache& query_cache);
+
+protected:
+    Fence CreateFence(u32 value, bool is_stubbed) override;
+    Fence CreateFence(GPUVAddr addr, u32 value, bool is_stubbed) override;
+    void QueueFence(Fence& fence) override;
+    bool IsFenceSignaled(Fence& fence) const override;
+    void WaitFence(Fence& fence) override;
+
+private:
+    const VKDevice& device;
+    VKScheduler& scheduler;
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
index 343999cf5..8332b42aa 100644
--- a/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
+++ b/src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp
@@ -148,6 +148,8 @@ std::vector<vk::ShaderModule> VKGraphicsPipeline::CreateShaderModules(
             continue;
         }
 
+        device.SaveShader(stage->code);
+
         ci.codeSize = stage->code.size() * sizeof(u32);
         ci.pCode = stage->code.data();
         modules.push_back(device.GetLogical().CreateShaderModule(ci));
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
index 8fdc6400d..91b1b16a5 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.cpp
@@ -207,7 +207,7 @@ std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
         const GPUVAddr program_addr{GetShaderAddress(system, program)};
         const std::optional cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
         ASSERT(cpu_addr);
-        auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr;
+        auto shader = cpu_addr ? TryGet(*cpu_addr) : null_shader;
         if (!shader) {
             const auto host_ptr{memory_manager.GetPointer(program_addr)};
 
@@ -218,7 +218,11 @@ std::array<Shader, Maxwell::MaxShaderProgram> VKPipelineCache::GetShaders() {
 
             shader = std::make_shared<CachedShader>(system, stage, program_addr, *cpu_addr,
                                                     std::move(code), stage_offset);
-            Register(shader);
+            if (cpu_addr) {
+                Register(shader);
+            } else {
+                null_shader = shader;
+            }
         }
         shaders[index] = std::move(shader);
     }
@@ -261,7 +265,7 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach
     const auto cpu_addr = memory_manager.GpuToCpuAddress(program_addr);
     ASSERT(cpu_addr);
 
-    auto shader = cpu_addr ? TryGet(*cpu_addr) : nullptr;
+    auto shader = cpu_addr ? TryGet(*cpu_addr) : null_kernel;
     if (!shader) {
         // No shader found - create a new one
         const auto host_ptr = memory_manager.GetPointer(program_addr);
@@ -271,7 +275,11 @@ VKComputePipeline& VKPipelineCache::GetComputePipeline(const ComputePipelineCach
         shader = std::make_shared<CachedShader>(system, Tegra::Engines::ShaderType::Compute,
                                                 program_addr, *cpu_addr, std::move(code),
                                                 kernel_main_offset);
-        Register(shader);
+        if (cpu_addr) {
+            Register(shader);
+        } else {
+            null_kernel = shader;
+        }
     }
 
     Specialization specialization;
@@ -330,8 +338,10 @@ VKPipelineCache::DecompileShaders(const GraphicsPipelineCacheKey& key) {
 
     Specialization specialization;
     if (fixed_state.rasterizer.Topology() == Maxwell::PrimitiveTopology::Points) {
-        ASSERT(fixed_state.rasterizer.point_size != 0);
-        std::memcpy(&specialization.point_size, &fixed_state.rasterizer.point_size, sizeof(u32));
+        float point_size;
+        std::memcpy(&point_size, &fixed_state.rasterizer.point_size, sizeof(float));
+        specialization.point_size = point_size;
+        ASSERT(point_size != 0.0f);
     }
     for (std::size_t i = 0; i < Maxwell::NumVertexAttributes; ++i) {
         specialization.attribute_types[i] = fixed_state.vertex_input.attributes[i].Type();
diff --git a/src/video_core/renderer_vulkan/vk_pipeline_cache.h b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
index 7ccdb7083..602a0a340 100644
--- a/src/video_core/renderer_vulkan/vk_pipeline_cache.h
+++ b/src/video_core/renderer_vulkan/vk_pipeline_cache.h
@@ -182,6 +182,9 @@ private:
     VKUpdateDescriptorQueue& update_descriptor_queue;
     VKRenderPassCache& renderpass_cache;
 
+    Shader null_shader{};
+    Shader null_kernel{};
+
     std::array<Shader, Maxwell::MaxShaderProgram> last_shaders;
 
     GraphicsPipelineCacheKey last_graphics_key;
diff --git a/src/video_core/renderer_vulkan/vk_query_cache.cpp b/src/video_core/renderer_vulkan/vk_query_cache.cpp
index 0966c7ff7..813f7c162 100644
--- a/src/video_core/renderer_vulkan/vk_query_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_query_cache.cpp
@@ -113,8 +113,19 @@ u64 HostCounter::BlockingQuery() const {
     if (ticks >= cache.Scheduler().Ticks()) {
         cache.Scheduler().Flush();
     }
-    return cache.Device().GetLogical().GetQueryResult<u64>(
-        query.first, query.second, VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT);
+    u64 data;
+    const VkResult result = cache.Device().GetLogical().GetQueryResults(
+        query.first, query.second, 1, sizeof(data), &data, sizeof(data),
+        VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT);
+    switch (result) {
+    case VK_SUCCESS:
+        return data;
+    case VK_ERROR_DEVICE_LOST:
+        cache.Device().ReportLoss();
+        [[fallthrough]];
+    default:
+        throw vk::Exception(result);
+    }
 }
 
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index 71007bbe8..68464e637 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -17,6 +17,7 @@
 #include "common/microprofile.h"
 #include "core/core.h"
 #include "core/memory.h"
+#include "core/settings.h"
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/renderer_vulkan/fixed_pipeline_state.h"
@@ -299,7 +300,9 @@ RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWind
       pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue,
                      renderpass_cache),
       buffer_cache(*this, system, device, memory_manager, scheduler, staging_pool),
-      sampler_cache(device), query_cache(system, *this, device, scheduler) {
+      sampler_cache(device),
+      fence_manager(system, *this, device, scheduler, texture_cache, buffer_cache, query_cache),
+      query_cache(system, *this, device, scheduler) {
     scheduler.SetQueryCache(query_cache);
 }
 
@@ -347,11 +350,6 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
 
     buffer_bindings.Bind(scheduler);
 
-    if (device.IsNvDeviceDiagnosticCheckpoints()) {
-        scheduler.Record(
-            [&pipeline](vk::CommandBuffer cmdbuf) { cmdbuf.SetCheckpointNV(&pipeline); });
-    }
-
     BeginTransformFeedback();
 
     const auto pipeline_layout = pipeline.GetLayout();
@@ -365,6 +363,8 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
     });
 
     EndTransformFeedback();
+
+    system.GPU().TickWork();
 }
 
 void RasterizerVulkan::Clear() {
@@ -478,11 +478,6 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
     TransitionImages(image_views, VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
                      VK_ACCESS_SHADER_READ_BIT | VK_ACCESS_SHADER_WRITE_BIT);
 
-    if (device.IsNvDeviceDiagnosticCheckpoints()) {
-        scheduler.Record(
-            [&pipeline](vk::CommandBuffer cmdbuf) { cmdbuf.SetCheckpointNV(nullptr); });
-    }
-
     scheduler.Record([grid_x = launch_desc.grid_dim_x, grid_y = launch_desc.grid_dim_y,
                       grid_z = launch_desc.grid_dim_z, pipeline_handle = pipeline.GetHandle(),
                       layout = pipeline.GetLayout(),
@@ -514,6 +509,13 @@ void RasterizerVulkan::FlushRegion(VAddr addr, u64 size) {
     query_cache.FlushRegion(addr, size);
 }
 
+bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size) {
+    if (!Settings::IsGPULevelHigh()) {
+        return buffer_cache.MustFlushRegion(addr, size);
+    }
+    return texture_cache.MustFlushRegion(addr, size) || buffer_cache.MustFlushRegion(addr, size);
+}
+
 void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) {
     if (addr == 0 || size == 0) {
         return;
@@ -524,6 +526,47 @@ void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size) {
     query_cache.InvalidateRegion(addr, size);
 }
 
+void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
+    if (addr == 0 || size == 0) {
+        return;
+    }
+    texture_cache.OnCPUWrite(addr, size);
+    pipeline_cache.InvalidateRegion(addr, size);
+    buffer_cache.OnCPUWrite(addr, size);
+    query_cache.InvalidateRegion(addr, size);
+}
+
+void RasterizerVulkan::SyncGuestHost() {
+    texture_cache.SyncGuestHost();
+    buffer_cache.SyncGuestHost();
+}
+
+void RasterizerVulkan::SignalSemaphore(GPUVAddr addr, u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        gpu.MemoryManager().Write<u32>(addr, value);
+        return;
+    }
+    fence_manager.SignalSemaphore(addr, value);
+}
+
+void RasterizerVulkan::SignalSyncPoint(u32 value) {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        gpu.IncrementSyncPoint(value);
+        return;
+    }
+    fence_manager.SignalSyncPoint(value);
+}
+
+void RasterizerVulkan::ReleaseFences() {
+    auto& gpu{system.GPU()};
+    if (!gpu.IsAsync()) {
+        return;
+    }
+    fence_manager.WaitPendingFences();
+}
+
 void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     FlushRegion(addr, size);
     InvalidateRegion(addr, size);
@@ -834,8 +877,12 @@ void RasterizerVulkan::SetupVertexArrays(FixedPipelineState::VertexInput& vertex
         const GPUVAddr start{vertex_array.StartAddress()};
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
 
-        ASSERT(end > start);
-        const std::size_t size{end - start + 1};
+        ASSERT(end >= start);
+        const std::size_t size{end - start};
+        if (size == 0) {
+            buffer_bindings.AddVertexBinding(DefaultBuffer(), 0);
+            continue;
+        }
         const auto [buffer, offset] = buffer_cache.UploadMemory(start, size);
         buffer_bindings.AddVertexBinding(buffer, offset);
     }
@@ -990,8 +1037,7 @@ void RasterizerVulkan::SetupConstBuffer(const ConstBufferEntry& entry,
                                         const Tegra::Engines::ConstBufferInfo& buffer) {
     if (!buffer.enabled) {
         // Set values to zero to unbind buffers
-        update_descriptor_queue.AddBuffer(buffer_cache.GetEmptyBuffer(sizeof(float)), 0,
-                                          sizeof(float));
+        update_descriptor_queue.AddBuffer(DefaultBuffer(), 0, DEFAULT_BUFFER_SIZE);
         return;
     }
 
@@ -1014,7 +1060,9 @@ void RasterizerVulkan::SetupGlobalBuffer(const GlobalBufferEntry& entry, GPUVAdd
     if (size == 0) {
         // Sometimes global memory pointers don't have a proper size. Upload a dummy entry
         // because Vulkan doesn't like empty buffers.
-        constexpr std::size_t dummy_size = 4;
+        // Note: Do *not* use DefaultBuffer() here, storage buffers can be written breaking the
+        // default buffer.
+        static constexpr std::size_t dummy_size = 4;
         const auto buffer = buffer_cache.GetEmptyBuffer(dummy_size);
         update_descriptor_queue.AddBuffer(buffer, 0, dummy_size);
         return;
@@ -1179,7 +1227,7 @@ std::size_t RasterizerVulkan::CalculateVertexArraysSize() const {
         const GPUVAddr end{regs.vertex_array_limit[index].LimitAddress()};
         DEBUG_ASSERT(end >= start);
 
-        size += (end - start + 1) * regs.vertex_array[index].enable;
+        size += (end - start) * regs.vertex_array[index].enable;
     }
     return size;
 }
@@ -1226,4 +1274,29 @@ RenderPassParams RasterizerVulkan::GetRenderPassParams(Texceptions texceptions)
     return renderpass_params;
 }
 
+VkBuffer RasterizerVulkan::DefaultBuffer() {
+    if (default_buffer) {
+        return *default_buffer;
+    }
+
+    VkBufferCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    ci.size = DEFAULT_BUFFER_SIZE;
+    ci.usage = VK_BUFFER_USAGE_TRANSFER_DST_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT |
+               VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT;
+    ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
+    ci.queueFamilyIndexCount = 0;
+    ci.pQueueFamilyIndices = nullptr;
+    default_buffer = device.GetLogical().CreateBuffer(ci);
+    default_buffer_commit = memory_manager.Commit(default_buffer, false);
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([buffer = *default_buffer](vk::CommandBuffer cmdbuf) {
+        cmdbuf.FillBuffer(buffer, 0, DEFAULT_BUFFER_SIZE, 0);
+    });
+    return *default_buffer;
+}
+
 } // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index d9108f862..d41a7929e 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -21,6 +21,7 @@
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_fence_manager.h"
 #include "video_core/renderer_vulkan/vk_memory_manager.h"
 #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
 #include "video_core/renderer_vulkan/vk_query_cache.h"
@@ -118,7 +119,13 @@ public:
     void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
     void FlushAll() override;
     void FlushRegion(VAddr addr, u64 size) override;
+    bool MustFlushRegion(VAddr addr, u64 size) override;
     void InvalidateRegion(VAddr addr, u64 size) override;
+    void OnCPUWrite(VAddr addr, u64 size) override;
+    void SyncGuestHost() override;
+    void SignalSemaphore(GPUVAddr addr, u32 value) override;
+    void SignalSyncPoint(u32 value) override;
+    void ReleaseFences() override;
     void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
     void FlushCommands() override;
     void TickFrame() override;
@@ -148,6 +155,7 @@ private:
     using Texceptions = std::bitset<Maxwell::NumRenderTargets + 1>;
 
     static constexpr std::size_t ZETA_TEXCEPTION_INDEX = 8;
+    static constexpr VkDeviceSize DEFAULT_BUFFER_SIZE = 4 * sizeof(float);
 
     void FlushWork();
 
@@ -240,6 +248,8 @@ private:
 
     RenderPassParams GetRenderPassParams(Texceptions texceptions) const;
 
+    VkBuffer DefaultBuffer();
+
     Core::System& system;
     Core::Frontend::EmuWindow& render_window;
     VKScreenInfo& screen_info;
@@ -261,8 +271,12 @@ private:
     VKPipelineCache pipeline_cache;
     VKBufferCache buffer_cache;
     VKSamplerCache sampler_cache;
+    VKFenceManager fence_manager;
     VKQueryCache query_cache;
 
+    vk::Buffer default_buffer;
+    VKMemoryCommit default_buffer_commit;
+
     std::array<View, Maxwell::NumRenderTargets> color_attachments;
     View zeta_attachment;
 
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp
index 900f551b3..ae7ba3eb5 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.cpp
+++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp
@@ -166,7 +166,15 @@ void VKScheduler::SubmitExecution(VkSemaphore semaphore) {
     submit_info.pCommandBuffers = current_cmdbuf.address();
     submit_info.signalSemaphoreCount = semaphore ? 1 : 0;
     submit_info.pSignalSemaphores = &semaphore;
-    device.GetGraphicsQueue().Submit(submit_info, *current_fence);
+    switch (const VkResult result = device.GetGraphicsQueue().Submit(submit_info, *current_fence)) {
+    case VK_SUCCESS:
+        break;
+    case VK_ERROR_DEVICE_LOST:
+        device.ReportLoss();
+        [[fallthrough]];
+    default:
+        vk::Check(result);
+    }
 }
 
 void VKScheduler::AllocateNewContext() {
diff --git a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
index abd836efe..45c180221 100644
--- a/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
+++ b/src/video_core/renderer_vulkan/vk_staging_buffer_pool.cpp
@@ -78,7 +78,7 @@ VKBuffer& VKStagingBufferPool::CreateStagingBuffer(std::size_t size, bool host_v
     ci.size = 1ULL << log2;
     ci.usage = VK_BUFFER_USAGE_TRANSFER_SRC_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT |
                VK_BUFFER_USAGE_UNIFORM_BUFFER_BIT | VK_BUFFER_USAGE_STORAGE_BUFFER_BIT |
-               VK_BUFFER_USAGE_INDEX_BUFFER_BIT;
+               VK_BUFFER_USAGE_INDEX_BUFFER_BIT | VK_BUFFER_USAGE_VERTEX_BUFFER_BIT;
     ci.sharingMode = VK_SHARING_MODE_EXCLUSIVE;
     ci.queueFamilyIndexCount = 0;
     ci.pQueueFamilyIndices = nullptr;
diff --git a/src/video_core/renderer_vulkan/wrapper.cpp b/src/video_core/renderer_vulkan/wrapper.cpp
index 9b94dfff1..7f5bc1404 100644
--- a/src/video_core/renderer_vulkan/wrapper.cpp
+++ b/src/video_core/renderer_vulkan/wrapper.cpp
@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <exception>
 #include <memory>
 #include <optional>
@@ -16,6 +17,23 @@ namespace Vulkan::vk {
 
 namespace {
 
+void SortPhysicalDevices(std::vector<VkPhysicalDevice>& devices, const InstanceDispatch& dld) {
+    std::stable_sort(devices.begin(), devices.end(), [&](auto lhs, auto rhs) {
+        // This will call Vulkan more than needed, but these calls are cheap.
+        const auto lhs_properties = vk::PhysicalDevice(lhs, dld).GetProperties();
+        const auto rhs_properties = vk::PhysicalDevice(rhs, dld).GetProperties();
+
+        // Prefer discrete GPUs, Nvidia over AMD, AMD over Intel, Intel over the rest.
+        const bool preferred =
+            (lhs_properties.deviceType == VK_PHYSICAL_DEVICE_TYPE_DISCRETE_GPU &&
+             rhs_properties.deviceType != VK_PHYSICAL_DEVICE_TYPE_DISCRETE_GPU) ||
+            (lhs_properties.vendorID == 0x10DE && rhs_properties.vendorID != 0x10DE) ||
+            (lhs_properties.vendorID == 0x1002 && rhs_properties.vendorID != 0x1002) ||
+            (lhs_properties.vendorID == 0x8086 && rhs_properties.vendorID != 0x8086);
+        return !preferred;
+    });
+}
+
 template <typename T>
 bool Proc(T& result, const InstanceDispatch& dld, const char* proc_name,
           VkInstance instance = nullptr) noexcept {
@@ -61,9 +79,9 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkCmdPipelineBarrier);
     X(vkCmdPushConstants);
     X(vkCmdSetBlendConstants);
-    X(vkCmdSetCheckpointNV);
     X(vkCmdSetDepthBias);
     X(vkCmdSetDepthBounds);
+    X(vkCmdSetEvent);
     X(vkCmdSetScissor);
     X(vkCmdSetStencilCompareMask);
     X(vkCmdSetStencilReference);
@@ -76,6 +94,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkCreateDescriptorPool);
     X(vkCreateDescriptorSetLayout);
     X(vkCreateDescriptorUpdateTemplateKHR);
+    X(vkCreateEvent);
     X(vkCreateFence);
     X(vkCreateFramebuffer);
     X(vkCreateGraphicsPipelines);
@@ -94,6 +113,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkDestroyDescriptorPool);
     X(vkDestroyDescriptorSetLayout);
     X(vkDestroyDescriptorUpdateTemplateKHR);
+    X(vkDestroyEvent);
     X(vkDestroyFence);
     X(vkDestroyFramebuffer);
     X(vkDestroyImage);
@@ -113,10 +133,10 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkFreeMemory);
     X(vkGetBufferMemoryRequirements);
     X(vkGetDeviceQueue);
+    X(vkGetEventStatus);
     X(vkGetFenceStatus);
     X(vkGetImageMemoryRequirements);
     X(vkGetQueryPoolResults);
-    X(vkGetQueueCheckpointDataNV);
     X(vkMapMemory);
     X(vkQueueSubmit);
     X(vkResetFences);
@@ -271,6 +291,10 @@ void Destroy(VkDevice device, VkDeviceMemory handle, const DeviceDispatch& dld)
     dld.vkFreeMemory(device, handle, nullptr);
 }
 
+void Destroy(VkDevice device, VkEvent handle, const DeviceDispatch& dld) noexcept {
+    dld.vkDestroyEvent(device, handle, nullptr);
+}
+
 void Destroy(VkDevice device, VkFence handle, const DeviceDispatch& dld) noexcept {
     dld.vkDestroyFence(device, handle, nullptr);
 }
@@ -383,7 +407,8 @@ std::optional<std::vector<VkPhysicalDevice>> Instance::EnumeratePhysicalDevices(
     if (dld->vkEnumeratePhysicalDevices(handle, &num, physical_devices.data()) != VK_SUCCESS) {
         return std::nullopt;
     }
-    return physical_devices;
+    SortPhysicalDevices(physical_devices, *dld);
+    return std::make_optional(std::move(physical_devices));
 }
 
 DebugCallback Instance::TryCreateDebugCallback(
@@ -409,17 +434,6 @@ DebugCallback Instance::TryCreateDebugCallback(
     return DebugCallback(messenger, handle, *dld);
 }
 
-std::vector<VkCheckpointDataNV> Queue::GetCheckpointDataNV(const DeviceDispatch& dld) const {
-    if (!dld.vkGetQueueCheckpointDataNV) {
-        return {};
-    }
-    u32 num;
-    dld.vkGetQueueCheckpointDataNV(queue, &num, nullptr);
-    std::vector<VkCheckpointDataNV> checkpoints(num);
-    dld.vkGetQueueCheckpointDataNV(queue, &num, checkpoints.data());
-    return checkpoints;
-}
-
 void Buffer::BindMemory(VkDeviceMemory memory, VkDeviceSize offset) const {
     Check(dld->vkBindBufferMemory(owner, handle, memory, offset));
 }
@@ -469,12 +483,11 @@ std::vector<VkImage> SwapchainKHR::GetImages() const {
 }
 
 Device Device::Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
-                      Span<const char*> enabled_extensions,
-                      const VkPhysicalDeviceFeatures2& enabled_features,
+                      Span<const char*> enabled_extensions, const void* next,
                       DeviceDispatch& dld) noexcept {
     VkDeviceCreateInfo ci;
     ci.sType = VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO;
-    ci.pNext = &enabled_features;
+    ci.pNext = next;
     ci.flags = 0;
     ci.queueCreateInfoCount = queues_ci.size();
     ci.pQueueCreateInfos = queues_ci.data();
@@ -613,6 +626,16 @@ ShaderModule Device::CreateShaderModule(const VkShaderModuleCreateInfo& ci) cons
     return ShaderModule(object, handle, *dld);
 }
 
+Event Device::CreateEvent() const {
+    VkEventCreateInfo ci;
+    ci.sType = VK_STRUCTURE_TYPE_EVENT_CREATE_INFO;
+    ci.pNext = nullptr;
+    ci.flags = 0;
+    VkEvent object;
+    Check(dld->vkCreateEvent(handle, &ci, nullptr, &object));
+    return Event(object, handle, *dld);
+}
+
 SwapchainKHR Device::CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const {
     VkSwapchainKHR object;
     Check(dld->vkCreateSwapchainKHR(handle, &ci, nullptr, &object));
diff --git a/src/video_core/renderer_vulkan/wrapper.h b/src/video_core/renderer_vulkan/wrapper.h
index fb3657819..bda16a2cb 100644
--- a/src/video_core/renderer_vulkan/wrapper.h
+++ b/src/video_core/renderer_vulkan/wrapper.h
@@ -197,9 +197,9 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkCmdPipelineBarrier vkCmdPipelineBarrier;
     PFN_vkCmdPushConstants vkCmdPushConstants;
     PFN_vkCmdSetBlendConstants vkCmdSetBlendConstants;
-    PFN_vkCmdSetCheckpointNV vkCmdSetCheckpointNV;
     PFN_vkCmdSetDepthBias vkCmdSetDepthBias;
     PFN_vkCmdSetDepthBounds vkCmdSetDepthBounds;
+    PFN_vkCmdSetEvent vkCmdSetEvent;
     PFN_vkCmdSetScissor vkCmdSetScissor;
     PFN_vkCmdSetStencilCompareMask vkCmdSetStencilCompareMask;
     PFN_vkCmdSetStencilReference vkCmdSetStencilReference;
@@ -212,6 +212,7 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkCreateDescriptorPool vkCreateDescriptorPool;
     PFN_vkCreateDescriptorSetLayout vkCreateDescriptorSetLayout;
     PFN_vkCreateDescriptorUpdateTemplateKHR vkCreateDescriptorUpdateTemplateKHR;
+    PFN_vkCreateEvent vkCreateEvent;
     PFN_vkCreateFence vkCreateFence;
     PFN_vkCreateFramebuffer vkCreateFramebuffer;
     PFN_vkCreateGraphicsPipelines vkCreateGraphicsPipelines;
@@ -230,6 +231,7 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkDestroyDescriptorPool vkDestroyDescriptorPool;
     PFN_vkDestroyDescriptorSetLayout vkDestroyDescriptorSetLayout;
     PFN_vkDestroyDescriptorUpdateTemplateKHR vkDestroyDescriptorUpdateTemplateKHR;
+    PFN_vkDestroyEvent vkDestroyEvent;
     PFN_vkDestroyFence vkDestroyFence;
     PFN_vkDestroyFramebuffer vkDestroyFramebuffer;
     PFN_vkDestroyImage vkDestroyImage;
@@ -249,10 +251,10 @@ struct DeviceDispatch : public InstanceDispatch {
     PFN_vkFreeMemory vkFreeMemory;
     PFN_vkGetBufferMemoryRequirements vkGetBufferMemoryRequirements;
     PFN_vkGetDeviceQueue vkGetDeviceQueue;
+    PFN_vkGetEventStatus vkGetEventStatus;
     PFN_vkGetFenceStatus vkGetFenceStatus;
     PFN_vkGetImageMemoryRequirements vkGetImageMemoryRequirements;
     PFN_vkGetQueryPoolResults vkGetQueryPoolResults;
-    PFN_vkGetQueueCheckpointDataNV vkGetQueueCheckpointDataNV;
     PFN_vkMapMemory vkMapMemory;
     PFN_vkQueueSubmit vkQueueSubmit;
     PFN_vkResetFences vkResetFences;
@@ -281,6 +283,7 @@ void Destroy(VkDevice, VkDescriptorPool, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkDescriptorSetLayout, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkDescriptorUpdateTemplateKHR, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkDeviceMemory, const DeviceDispatch&) noexcept;
+void Destroy(VkDevice, VkEvent, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkFence, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkFramebuffer, const DeviceDispatch&) noexcept;
 void Destroy(VkDevice, VkImage, const DeviceDispatch&) noexcept;
@@ -567,12 +570,8 @@ public:
     /// Construct a queue handle.
     constexpr Queue(VkQueue queue, const DeviceDispatch& dld) noexcept : queue{queue}, dld{&dld} {}
 
-    /// Returns the checkpoint data.
-    /// @note Returns an empty vector when the function pointer is not present.
-    std::vector<VkCheckpointDataNV> GetCheckpointDataNV(const DeviceDispatch& dld) const;
-
-    void Submit(Span<VkSubmitInfo> submit_infos, VkFence fence) const {
-        Check(dld->vkQueueSubmit(queue, submit_infos.size(), submit_infos.data(), fence));
+    VkResult Submit(Span<VkSubmitInfo> submit_infos, VkFence fence) const noexcept {
+        return dld->vkQueueSubmit(queue, submit_infos.size(), submit_infos.data(), fence);
     }
 
     VkResult Present(const VkPresentInfoKHR& present_info) const noexcept {
@@ -654,13 +653,21 @@ public:
     std::vector<VkImage> GetImages() const;
 };
 
+class Event : public Handle<VkEvent, VkDevice, DeviceDispatch> {
+    using Handle<VkEvent, VkDevice, DeviceDispatch>::Handle;
+
+public:
+    VkResult GetStatus() const noexcept {
+        return dld->vkGetEventStatus(owner, handle);
+    }
+};
+
 class Device : public Handle<VkDevice, NoOwner, DeviceDispatch> {
     using Handle<VkDevice, NoOwner, DeviceDispatch>::Handle;
 
 public:
     static Device Create(VkPhysicalDevice physical_device, Span<VkDeviceQueueCreateInfo> queues_ci,
-                         Span<const char*> enabled_extensions,
-                         const VkPhysicalDeviceFeatures2& enabled_features,
+                         Span<const char*> enabled_extensions, const void* next,
                          DeviceDispatch& dld) noexcept;
 
     Queue GetQueue(u32 family_index) const noexcept;
@@ -702,6 +709,8 @@ public:
 
     ShaderModule CreateShaderModule(const VkShaderModuleCreateInfo& ci) const;
 
+    Event CreateEvent() const;
+
     SwapchainKHR CreateSwapchainKHR(const VkSwapchainCreateInfoKHR& ci) const;
 
     DeviceMemory TryAllocateMemory(const VkMemoryAllocateInfo& ai) const noexcept;
@@ -734,18 +743,11 @@ public:
         dld->vkResetQueryPoolEXT(handle, query_pool, first, count);
     }
 
-    void GetQueryResults(VkQueryPool query_pool, u32 first, u32 count, std::size_t data_size,
-                         void* data, VkDeviceSize stride, VkQueryResultFlags flags) const {
-        Check(dld->vkGetQueryPoolResults(handle, query_pool, first, count, data_size, data, stride,
-                                         flags));
-    }
-
-    template <typename T>
-    T GetQueryResult(VkQueryPool query_pool, u32 first, VkQueryResultFlags flags) const {
-        static_assert(std::is_trivially_copyable_v<T>);
-        T value;
-        GetQueryResults(query_pool, first, 1, sizeof(T), &value, sizeof(T), flags);
-        return value;
+    VkResult GetQueryResults(VkQueryPool query_pool, u32 first, u32 count, std::size_t data_size,
+                             void* data, VkDeviceSize stride, VkQueryResultFlags flags) const
+        noexcept {
+        return dld->vkGetQueryPoolResults(handle, query_pool, first, count, data_size, data, stride,
+                                          flags);
     }
 };
 
@@ -920,10 +922,6 @@ public:
         dld->vkCmdPushConstants(handle, layout, flags, offset, size, values);
     }
 
-    void SetCheckpointNV(const void* checkpoint_marker) const noexcept {
-        dld->vkCmdSetCheckpointNV(handle, checkpoint_marker);
-    }
-
     void SetViewport(u32 first, Span<VkViewport> viewports) const noexcept {
         dld->vkCmdSetViewport(handle, first, viewports.size(), viewports.data());
     }
@@ -956,6 +954,10 @@ public:
         dld->vkCmdSetDepthBounds(handle, min_depth_bounds, max_depth_bounds);
     }
 
+    void SetEvent(VkEvent event, VkPipelineStageFlags stage_flags) const noexcept {
+        dld->vkCmdSetEvent(handle, event, stage_flags);
+    }
+
     void BindTransformFeedbackBuffersEXT(u32 first, u32 count, const VkBuffer* buffers,
                                          const VkDeviceSize* offsets,
                                          const VkDeviceSize* sizes) const noexcept {
diff --git a/src/video_core/shader/control_flow.cpp b/src/video_core/shader/control_flow.cpp
index 6d313963a..e00a3fb70 100644
--- a/src/video_core/shader/control_flow.cpp
+++ b/src/video_core/shader/control_flow.cpp
@@ -587,8 +587,6 @@ bool TryQuery(CFGRebuildState& state) {
     return true;
 }
 
-} // Anonymous namespace
-
 void InsertBranch(ASTManager& mm, const BlockBranchInfo& branch_info) {
     const auto get_expr = ([&](const Condition& cond) -> Expr {
         Expr result{};
@@ -655,6 +653,8 @@ void DecompileShader(CFGRebuildState& state) {
     state.manager->Decompile();
 }
 
+} // Anonymous namespace
+
 std::unique_ptr<ShaderCharacteristics> ScanFlow(const ProgramCode& program_code, u32 start_address,
                                                 const CompilerSettings& settings,
                                                 Registry& registry) {
diff --git a/src/video_core/shader/decode/arithmetic_half.cpp b/src/video_core/shader/decode/arithmetic_half.cpp
index ee7d9a29d..a276aee44 100644
--- a/src/video_core/shader/decode/arithmetic_half.cpp
+++ b/src/video_core/shader/decode/arithmetic_half.cpp
@@ -19,22 +19,46 @@ u32 ShaderIR::DecodeArithmeticHalf(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
-    if (opcode->get().GetId() == OpCode::Id::HADD2_C ||
-        opcode->get().GetId() == OpCode::Id::HADD2_R) {
+    bool negate_a = false;
+    bool negate_b = false;
+    bool absolute_a = false;
+    bool absolute_b = false;
+
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::HADD2_R:
         if (instr.alu_half.ftz == 0) {
             LOG_DEBUG(HW_GPU, "{} without FTZ is not implemented", opcode->get().GetName());
         }
+        negate_a = ((instr.value >> 43) & 1) != 0;
+        negate_b = ((instr.value >> 31) & 1) != 0;
+        absolute_a = ((instr.value >> 44) & 1) != 0;
+        absolute_b = ((instr.value >> 30) & 1) != 0;
+        break;
+    case OpCode::Id::HADD2_C:
+        if (instr.alu_half.ftz == 0) {
+            LOG_DEBUG(HW_GPU, "{} without FTZ is not implemented", opcode->get().GetName());
+        }
+        negate_a = ((instr.value >> 43) & 1) != 0;
+        negate_b = ((instr.value >> 56) & 1) != 0;
+        absolute_a = ((instr.value >> 44) & 1) != 0;
+        absolute_b = ((instr.value >> 54) & 1) != 0;
+        break;
+    case OpCode::Id::HMUL2_R:
+        negate_a = ((instr.value >> 43) & 1) != 0;
+        absolute_a = ((instr.value >> 44) & 1) != 0;
+        absolute_b = ((instr.value >> 30) & 1) != 0;
+        break;
+    case OpCode::Id::HMUL2_C:
+        negate_b = ((instr.value >> 31) & 1) != 0;
+        absolute_a = ((instr.value >> 44) & 1) != 0;
+        absolute_b = ((instr.value >> 54) & 1) != 0;
+        break;
     }
 
-    const bool negate_a =
-        opcode->get().GetId() != OpCode::Id::HMUL2_R && instr.alu_half.negate_a != 0;
-    const bool negate_b =
-        opcode->get().GetId() != OpCode::Id::HMUL2_C && instr.alu_half.negate_b != 0;
-
     Node op_a = UnpackHalfFloat(GetRegister(instr.gpr8), instr.alu_half.type_a);
-    op_a = GetOperandAbsNegHalf(op_a, instr.alu_half.abs_a, negate_a);
+    op_a = GetOperandAbsNegHalf(op_a, absolute_a, negate_a);
 
-    auto [type_b, op_b] = [&]() -> std::tuple<HalfType, Node> {
+    auto [type_b, op_b] = [this, instr, opcode]() -> std::pair<HalfType, Node> {
         switch (opcode->get().GetId()) {
         case OpCode::Id::HADD2_C:
         case OpCode::Id::HMUL2_C:
@@ -48,17 +72,16 @@ u32 ShaderIR::DecodeArithmeticHalf(NodeBlock& bb, u32 pc) {
         }
     }();
     op_b = UnpackHalfFloat(op_b, type_b);
-    // redeclaration to avoid a bug in clang with reusing local bindings in lambdas
-    Node op_b_alt = GetOperandAbsNegHalf(op_b, instr.alu_half.abs_b, negate_b);
+    op_b = GetOperandAbsNegHalf(op_b, absolute_b, negate_b);
 
-    Node value = [&]() {
+    Node value = [this, opcode, op_a, op_b = op_b] {
         switch (opcode->get().GetId()) {
         case OpCode::Id::HADD2_C:
         case OpCode::Id::HADD2_R:
-            return Operation(OperationCode::HAdd, PRECISE, op_a, op_b_alt);
+            return Operation(OperationCode::HAdd, PRECISE, op_a, op_b);
         case OpCode::Id::HMUL2_C:
         case OpCode::Id::HMUL2_R:
-            return Operation(OperationCode::HMul, PRECISE, op_a, op_b_alt);
+            return Operation(OperationCode::HMul, PRECISE, op_a, op_b);
         default:
             UNIMPLEMENTED_MSG("Unhandled half float instruction: {}", opcode->get().GetName());
             return Immediate(0);
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index 0f4c3103a..9af8c606d 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -249,8 +249,8 @@ u32 ShaderIR::DecodeArithmeticInteger(NodeBlock& bb, u32 pc) {
             }
             case OpCode::Id::LEA_IMM: {
                 const bool neg = instr.lea.imm.neg != 0;
-                return {Immediate(static_cast<u32>(instr.lea.imm.entry_a)),
-                        GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
+                return {GetOperandAbsNegInteger(GetRegister(instr.gpr8), false, neg, true),
+                        Immediate(static_cast<u32>(instr.lea.imm.entry_a)),
                         Immediate(static_cast<u32>(instr.lea.imm.entry_b))};
             }
             case OpCode::Id::LEA_RZ: {
diff --git a/src/video_core/texture_cache/surface_base.h b/src/video_core/texture_cache/surface_base.h
index c5ab21f56..79e10ffbb 100644
--- a/src/video_core/texture_cache/surface_base.h
+++ b/src/video_core/texture_cache/surface_base.h
@@ -192,6 +192,22 @@ public:
         index = index_;
     }
 
+    void SetMemoryMarked(bool is_memory_marked_) {
+        is_memory_marked = is_memory_marked_;
+    }
+
+    bool IsMemoryMarked() const {
+        return is_memory_marked;
+    }
+
+    void SetSyncPending(bool is_sync_pending_) {
+        is_sync_pending = is_sync_pending_;
+    }
+
+    bool IsSyncPending() const {
+        return is_sync_pending;
+    }
+
     void MarkAsPicked(bool is_picked_) {
         is_picked = is_picked_;
     }
@@ -303,6 +319,8 @@ private:
     bool is_target{};
     bool is_registered{};
     bool is_picked{};
+    bool is_memory_marked{};
+    bool is_sync_pending{};
     u32 index{NO_RT};
     u64 modification_tick{};
 };
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index 69ca08fd1..cf6bd005a 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -6,6 +6,7 @@
 
 #include <algorithm>
 #include <array>
+#include <list>
 #include <memory>
 #include <mutex>
 #include <set>
@@ -62,6 +63,30 @@ public:
         }
     }
 
+    void OnCPUWrite(VAddr addr, std::size_t size) {
+        std::lock_guard lock{mutex};
+
+        for (const auto& surface : GetSurfacesInRegion(addr, size)) {
+            if (surface->IsMemoryMarked()) {
+                UnmarkMemory(surface);
+                surface->SetSyncPending(true);
+                marked_for_unregister.emplace_back(surface);
+            }
+        }
+    }
+
+    void SyncGuestHost() {
+        std::lock_guard lock{mutex};
+
+        for (const auto& surface : marked_for_unregister) {
+            if (surface->IsRegistered()) {
+                surface->SetSyncPending(false);
+                Unregister(surface);
+            }
+        }
+        marked_for_unregister.clear();
+    }
+
     /**
      * Guarantees that rendertargets don't unregister themselves if the
      * collide. Protection is currently only done on 3D slices.
@@ -85,10 +110,20 @@ public:
             return a->GetModificationTick() < b->GetModificationTick();
         });
         for (const auto& surface : surfaces) {
+            mutex.unlock();
             FlushSurface(surface);
+            mutex.lock();
         }
     }
 
+    bool MustFlushRegion(VAddr addr, std::size_t size) {
+        std::lock_guard lock{mutex};
+
+        const auto surfaces = GetSurfacesInRegion(addr, size);
+        return std::any_of(surfaces.cbegin(), surfaces.cend(),
+                           [](const TSurface& surface) { return surface->IsModified(); });
+    }
+
     TView GetTextureSurface(const Tegra::Texture::TICEntry& tic,
                             const VideoCommon::Shader::Sampler& entry) {
         std::lock_guard lock{mutex};
@@ -206,8 +241,14 @@ public:
 
         auto surface_view = GetSurface(gpu_addr, *cpu_addr,
                                        SurfaceParams::CreateForFramebuffer(system, index), true);
-        if (render_targets[index].target)
-            render_targets[index].target->MarkAsRenderTarget(false, NO_RT);
+        if (render_targets[index].target) {
+            auto& surface = render_targets[index].target;
+            surface->MarkAsRenderTarget(false, NO_RT);
+            const auto& cr_params = surface->GetSurfaceParams();
+            if (!cr_params.is_tiled && Settings::values.use_asynchronous_gpu_emulation) {
+                AsyncFlushSurface(surface);
+            }
+        }
         render_targets[index].target = surface_view.first;
         render_targets[index].view = surface_view.second;
         if (render_targets[index].target)
@@ -284,6 +325,34 @@ public:
         return ++ticks;
     }
 
+    void CommitAsyncFlushes() {
+        committed_flushes.push_back(uncommitted_flushes);
+        uncommitted_flushes.reset();
+    }
+
+    bool HasUncommittedFlushes() const {
+        return uncommitted_flushes != nullptr;
+    }
+
+    bool ShouldWaitAsyncFlushes() const {
+        return !committed_flushes.empty() && committed_flushes.front() != nullptr;
+    }
+
+    void PopAsyncFlushes() {
+        if (committed_flushes.empty()) {
+            return;
+        }
+        auto& flush_list = committed_flushes.front();
+        if (!flush_list) {
+            committed_flushes.pop_front();
+            return;
+        }
+        for (TSurface& surface : *flush_list) {
+            FlushSurface(surface);
+        }
+        committed_flushes.pop_front();
+    }
+
 protected:
     explicit TextureCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
                           bool is_astc_supported)
@@ -345,9 +414,20 @@ protected:
         surface->SetCpuAddr(*cpu_addr);
         RegisterInnerCache(surface);
         surface->MarkAsRegistered(true);
+        surface->SetMemoryMarked(true);
         rasterizer.UpdatePagesCachedCount(*cpu_addr, size, 1);
     }
 
+    void UnmarkMemory(TSurface surface) {
+        if (!surface->IsMemoryMarked()) {
+            return;
+        }
+        const std::size_t size = surface->GetSizeInBytes();
+        const VAddr cpu_addr = surface->GetCpuAddr();
+        rasterizer.UpdatePagesCachedCount(cpu_addr, size, -1);
+        surface->SetMemoryMarked(false);
+    }
+
     void Unregister(TSurface surface) {
         if (guard_render_targets && surface->IsProtected()) {
             return;
@@ -355,9 +435,11 @@ protected:
         if (!guard_render_targets && surface->IsRenderTarget()) {
             ManageRenderTargetUnregister(surface);
         }
-        const std::size_t size = surface->GetSizeInBytes();
-        const VAddr cpu_addr = surface->GetCpuAddr();
-        rasterizer.UpdatePagesCachedCount(cpu_addr, size, -1);
+        UnmarkMemory(surface);
+        if (surface->IsSyncPending()) {
+            marked_for_unregister.remove(surface);
+            surface->SetSyncPending(false);
+        }
         UnregisterInnerCache(surface);
         surface->MarkAsRegistered(false);
         ReserveSurface(surface->GetSurfaceParams(), surface);
@@ -417,7 +499,7 @@ private:
      **/
     RecycleStrategy PickStrategy(std::vector<TSurface>& overlaps, const SurfaceParams& params,
                                  const GPUVAddr gpu_addr, const MatchTopologyResult untopological) {
-        if (Settings::values.use_accurate_gpu_emulation) {
+        if (Settings::IsGPULevelExtreme()) {
             return RecycleStrategy::Flush;
         }
         // 3D Textures decision
@@ -461,7 +543,7 @@ private:
         }
         switch (PickStrategy(overlaps, params, gpu_addr, untopological)) {
         case RecycleStrategy::Ignore: {
-            return InitializeSurface(gpu_addr, params, Settings::values.use_accurate_gpu_emulation);
+            return InitializeSurface(gpu_addr, params, Settings::IsGPULevelExtreme());
         }
         case RecycleStrategy::Flush: {
             std::sort(overlaps.begin(), overlaps.end(),
@@ -509,7 +591,7 @@ private:
         }
         const auto& final_params = new_surface->GetSurfaceParams();
         if (cr_params.type != final_params.type) {
-            if (Settings::values.use_accurate_gpu_emulation) {
+            if (Settings::IsGPULevelExtreme()) {
                 BufferCopy(current_surface, new_surface);
             }
         } else {
@@ -598,7 +680,7 @@ private:
         if (passed_tests == 0) {
             return {};
             // In Accurate GPU all tests should pass, else we recycle
-        } else if (Settings::values.use_accurate_gpu_emulation && passed_tests != overlaps.size()) {
+        } else if (Settings::IsGPULevelExtreme() && passed_tests != overlaps.size()) {
             return {};
         }
         for (const auto& surface : overlaps) {
@@ -668,7 +750,7 @@ private:
             for (const auto& surface : overlaps) {
                 if (!surface->MatchTarget(params.target)) {
                     if (overlaps.size() == 1 && surface->GetCpuAddr() == cpu_addr) {
-                        if (Settings::values.use_accurate_gpu_emulation) {
+                        if (Settings::IsGPULevelExtreme()) {
                             return std::nullopt;
                         }
                         Unregister(surface);
@@ -1106,6 +1188,13 @@ private:
         TView view;
     };
 
+    void AsyncFlushSurface(TSurface& surface) {
+        if (!uncommitted_flushes) {
+            uncommitted_flushes = std::make_shared<std::list<TSurface>>();
+        }
+        uncommitted_flushes->push_back(surface);
+    }
+
     VideoCore::RasterizerInterface& rasterizer;
 
     FormatLookupTable format_lookup_table;
@@ -1150,6 +1239,11 @@ private:
     std::unordered_map<u32, TSurface> invalid_cache;
     std::vector<u8> invalid_memory;
 
+    std::list<TSurface> marked_for_unregister;
+
+    std::shared_ptr<std::list<TSurface>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::list<TSurface>>> committed_flushes;
+
     StagingCache staging_cache;
     std::recursive_mutex mutex;
 };
diff --git a/src/video_core/textures/decoders.cpp b/src/video_core/textures/decoders.cpp
index 7df5f1452..fae8638ec 100644
--- a/src/video_core/textures/decoders.cpp
+++ b/src/video_core/textures/decoders.cpp
@@ -11,6 +11,7 @@
 #include "video_core/textures/texture.h"
 
 namespace Tegra::Texture {
+namespace {
 
 /**
  * This table represents the internal swizzle of a gob,
@@ -174,6 +175,8 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool
     }
 }
 
+} // Anonymous namespace
+
 void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel,
                       u32 out_bytes_per_pixel, u8* const swizzled_data, u8* const unswizzled_data,
                       bool unswizzle, u32 block_height, u32 block_depth, u32 width_spacing) {
diff --git a/src/video_core/textures/decoders.h b/src/video_core/textures/decoders.h
index e5eac3f3b..9f2d6d308 100644
--- a/src/video_core/textures/decoders.h
+++ b/src/video_core/textures/decoders.h
@@ -56,8 +56,7 @@ void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32
                       u32 bytes_per_pixel, u8* swizzled_data, u8* unswizzled_data, u32 block_height,
                       u32 offset_x, u32 offset_y);
 
-void SwizzleKepler(const u32 width, const u32 height, const u32 dst_x, const u32 dst_y,
-                   const u32 block_height, const std::size_t copy_size, const u8* source_data,
-                   u8* swizzle_data);
+void SwizzleKepler(u32 width, u32 height, u32 dst_x, u32 dst_y, u32 block_height,
+                   std::size_t copy_size, const u8* source_data, u8* swizzle_data);
 
 } // namespace Tegra::Texture
diff --git a/src/yuzu/applets/profile_select.cpp b/src/yuzu/applets/profile_select.cpp
index 6aff38735..4bc8ee726 100644
--- a/src/yuzu/applets/profile_select.cpp
+++ b/src/yuzu/applets/profile_select.cpp
@@ -17,6 +17,7 @@
 #include "yuzu/applets/profile_select.h"
 #include "yuzu/main.h"
 
+namespace {
 QString FormatUserEntryText(const QString& username, Common::UUID uuid) {
     return QtProfileSelectionDialog::tr(
                "%1\n%2", "%1 is the profile username, %2 is the formatted UUID (e.g. "
@@ -41,6 +42,7 @@ QPixmap GetIcon(Common::UUID uuid) {
 
     return icon.scaled(64, 64, Qt::IgnoreAspectRatio, Qt::SmoothTransformation);
 }
+} // Anonymous namespace
 
 QtProfileSelectionDialog::QtProfileSelectionDialog(QWidget* parent)
     : QDialog(parent), profile_manager(std::make_unique<Service::Account::ProfileManager>()) {
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index 7f6dfac84..a44eed047 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -12,7 +12,6 @@
 #include "input_common/main.h"
 #include "input_common/udp/client.h"
 #include "yuzu/configuration/config.h"
-#include "yuzu/uisettings.h"
 
 Config::Config() {
     // TODO: Don't hardcode the path; let the frontend decide where to put the config files.
@@ -212,12 +211,13 @@ const std::array<int, Settings::NativeKeyboard::NumKeyboardMods> Config::default
 // This must be in alphabetical order according to action name as it must have the same order as
 // UISetting::values.shortcuts, which is alphabetically ordered.
 // clang-format off
-const std::array<UISettings::Shortcut, 15> default_hotkeys{{
+const std::array<UISettings::Shortcut, 15> Config::default_hotkeys{{
     {QStringLiteral("Capture Screenshot"),       QStringLiteral("Main Window"), {QStringLiteral("Ctrl+P"), Qt::ApplicationShortcut}},
+    {QStringLiteral("Change Docked Mode"),       QStringLiteral("Main Window"), {QStringLiteral("F10"), Qt::ApplicationShortcut}},
     {QStringLiteral("Continue/Pause Emulation"), QStringLiteral("Main Window"), {QStringLiteral("F4"), Qt::WindowShortcut}},
     {QStringLiteral("Decrease Speed Limit"),     QStringLiteral("Main Window"), {QStringLiteral("-"), Qt::ApplicationShortcut}},
-    {QStringLiteral("Exit yuzu"),                QStringLiteral("Main Window"), {QStringLiteral("Ctrl+Q"), Qt::WindowShortcut}},
     {QStringLiteral("Exit Fullscreen"),          QStringLiteral("Main Window"), {QStringLiteral("Esc"), Qt::WindowShortcut}},
+    {QStringLiteral("Exit yuzu"),                QStringLiteral("Main Window"), {QStringLiteral("Ctrl+Q"), Qt::WindowShortcut}},
     {QStringLiteral("Fullscreen"),               QStringLiteral("Main Window"), {QStringLiteral("F11"), Qt::WindowShortcut}},
     {QStringLiteral("Increase Speed Limit"),     QStringLiteral("Main Window"), {QStringLiteral("+"), Qt::ApplicationShortcut}},
     {QStringLiteral("Load Amiibo"),              QStringLiteral("Main Window"), {QStringLiteral("F2"), Qt::ApplicationShortcut}},
@@ -227,7 +227,6 @@ const std::array<UISettings::Shortcut, 15> default_hotkeys{{
     {QStringLiteral("Toggle Filter Bar"),        QStringLiteral("Main Window"), {QStringLiteral("Ctrl+F"), Qt::WindowShortcut}},
     {QStringLiteral("Toggle Speed Limit"),       QStringLiteral("Main Window"), {QStringLiteral("Ctrl+Z"), Qt::ApplicationShortcut}},
     {QStringLiteral("Toggle Status Bar"),        QStringLiteral("Main Window"), {QStringLiteral("Ctrl+S"), Qt::WindowShortcut}},
-    {QStringLiteral("Change Docked Mode"),       QStringLiteral("Main Window"), {QStringLiteral("F10"), Qt::ApplicationShortcut}},
 }};
 // clang-format on
 
@@ -639,11 +638,13 @@ void Config::ReadRendererValues() {
     Settings::values.frame_limit = ReadSetting(QStringLiteral("frame_limit"), 100).toInt();
     Settings::values.use_disk_shader_cache =
         ReadSetting(QStringLiteral("use_disk_shader_cache"), true).toBool();
-    Settings::values.use_accurate_gpu_emulation =
-        ReadSetting(QStringLiteral("use_accurate_gpu_emulation"), false).toBool();
+    const int gpu_accuracy_level = ReadSetting(QStringLiteral("gpu_accuracy"), 0).toInt();
+    Settings::values.gpu_accuracy = static_cast<Settings::GPUAccuracy>(gpu_accuracy_level);
     Settings::values.use_asynchronous_gpu_emulation =
         ReadSetting(QStringLiteral("use_asynchronous_gpu_emulation"), false).toBool();
     Settings::values.use_vsync = ReadSetting(QStringLiteral("use_vsync"), true).toBool();
+    Settings::values.use_fast_gpu_time =
+        ReadSetting(QStringLiteral("use_fast_gpu_time"), true).toBool();
     Settings::values.force_30fps_mode =
         ReadSetting(QStringLiteral("force_30fps_mode"), false).toBool();
 
@@ -1080,11 +1081,12 @@ void Config::SaveRendererValues() {
     WriteSetting(QStringLiteral("frame_limit"), Settings::values.frame_limit, 100);
     WriteSetting(QStringLiteral("use_disk_shader_cache"), Settings::values.use_disk_shader_cache,
                  true);
-    WriteSetting(QStringLiteral("use_accurate_gpu_emulation"),
-                 Settings::values.use_accurate_gpu_emulation, false);
+    WriteSetting(QStringLiteral("gpu_accuracy"), static_cast<int>(Settings::values.gpu_accuracy),
+                 0);
     WriteSetting(QStringLiteral("use_asynchronous_gpu_emulation"),
                  Settings::values.use_asynchronous_gpu_emulation, false);
     WriteSetting(QStringLiteral("use_vsync"), Settings::values.use_vsync, true);
+    WriteSetting(QStringLiteral("use_fast_gpu_time"), Settings::values.use_fast_gpu_time, true);
     WriteSetting(QStringLiteral("force_30fps_mode"), Settings::values.force_30fps_mode, false);
 
     // Cast to double because Qt's written float values are not human-readable
diff --git a/src/yuzu/configuration/config.h b/src/yuzu/configuration/config.h
index ba6888004..5cd2a5feb 100644
--- a/src/yuzu/configuration/config.h
+++ b/src/yuzu/configuration/config.h
@@ -9,6 +9,7 @@
 #include <string>
 #include <QVariant>
 #include "core/settings.h"
+#include "yuzu/uisettings.h"
 
 class QSettings;
 
@@ -26,6 +27,7 @@ public:
         default_mouse_buttons;
     static const std::array<int, Settings::NativeKeyboard::NumKeyboardKeys> default_keyboard_keys;
     static const std::array<int, Settings::NativeKeyboard::NumKeyboardMods> default_keyboard_mods;
+    static const std::array<UISettings::Shortcut, 15> default_hotkeys;
 
 private:
     void ReadValues();
diff --git a/src/yuzu/configuration/configure_filesystem.cpp b/src/yuzu/configuration/configure_filesystem.cpp
index 29f540eb7..835ee821c 100644
--- a/src/yuzu/configuration/configure_filesystem.cpp
+++ b/src/yuzu/configuration/configure_filesystem.cpp
@@ -138,7 +138,7 @@ void ConfigureFilesystem::SetDirectory(DirectoryTarget target, QLineEdit* edit)
         str = QFileDialog::getOpenFileName(this, caption, QFileInfo(edit->text()).dir().path(),
                                            QStringLiteral("NX Gamecard;*.xci"));
     } else {
-        str = QFileDialog::getExistingDirectory(this, caption, edit->text());
+        str = QFileDialog::getExistingDirectory(this, caption, edit->text()) + QDir::separator();
     }
 
     if (str.isEmpty())
diff --git a/src/yuzu/configuration/configure_graphics_advanced.cpp b/src/yuzu/configuration/configure_graphics_advanced.cpp
index b9f429f84..5bb2ae555 100644
--- a/src/yuzu/configuration/configure_graphics_advanced.cpp
+++ b/src/yuzu/configuration/configure_graphics_advanced.cpp
@@ -19,9 +19,10 @@ ConfigureGraphicsAdvanced::~ConfigureGraphicsAdvanced() = default;
 
 void ConfigureGraphicsAdvanced::SetConfiguration() {
     const bool runtime_lock = !Core::System::GetInstance().IsPoweredOn();
-    ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
+    ui->gpu_accuracy->setCurrentIndex(static_cast<int>(Settings::values.gpu_accuracy));
     ui->use_vsync->setEnabled(runtime_lock);
     ui->use_vsync->setChecked(Settings::values.use_vsync);
+    ui->use_fast_gpu_time->setChecked(Settings::values.use_fast_gpu_time);
     ui->force_30fps_mode->setEnabled(runtime_lock);
     ui->force_30fps_mode->setChecked(Settings::values.force_30fps_mode);
     ui->anisotropic_filtering_combobox->setEnabled(runtime_lock);
@@ -29,8 +30,10 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
 }
 
 void ConfigureGraphicsAdvanced::ApplyConfiguration() {
-    Settings::values.use_accurate_gpu_emulation = ui->use_accurate_gpu_emulation->isChecked();
+    auto gpu_accuracy = static_cast<Settings::GPUAccuracy>(ui->gpu_accuracy->currentIndex());
+    Settings::values.gpu_accuracy = gpu_accuracy;
     Settings::values.use_vsync = ui->use_vsync->isChecked();
+    Settings::values.use_fast_gpu_time = ui->use_fast_gpu_time->isChecked();
     Settings::values.force_30fps_mode = ui->force_30fps_mode->isChecked();
     Settings::values.max_anisotropy = ui->anisotropic_filtering_combobox->currentIndex();
 }
diff --git a/src/yuzu/configuration/configure_graphics_advanced.ui b/src/yuzu/configuration/configure_graphics_advanced.ui
index 42eec278e..770b80c50 100644
--- a/src/yuzu/configuration/configure_graphics_advanced.ui
+++ b/src/yuzu/configuration/configure_graphics_advanced.ui
@@ -23,11 +23,34 @@
        </property>
        <layout class="QVBoxLayout" name="verticalLayout_3">
         <item>
-         <widget class="QCheckBox" name="use_accurate_gpu_emulation">
-          <property name="text">
-           <string>Use accurate GPU emulation (slow)</string>
-          </property>
-         </widget>
+         <layout class="QHBoxLayout" name="horizontalLayout_2">
+          <item>
+           <widget class="QLabel" name="label_gpu_accuracy">
+            <property name="text">
+             <string>Accuracy Level:</string>
+            </property>
+           </widget>
+          </item>
+          <item>
+           <widget class="QComboBox" name="gpu_accuracy">
+            <item>
+             <property name="text">
+              <string notr="true">Normal</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string notr="true">High</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string notr="true">Extreme(very slow)</string>
+             </property>
+            </item>
+           </widget>
+          </item>
+         </layout>
         </item>
         <item>
          <widget class="QCheckBox" name="use_vsync">
@@ -47,6 +70,13 @@
          </widget>
         </item>
         <item>
+         <widget class="QCheckBox" name="use_fast_gpu_time">
+          <property name="text">
+           <string>Use Fast GPU Time</string>
+          </property>
+         </widget>
+        </item>
+        <item>
          <layout class="QHBoxLayout" name="horizontalLayout_1">
           <item>
            <widget class="QLabel" name="af_label">
diff --git a/src/yuzu/configuration/configure_hotkeys.cpp b/src/yuzu/configuration/configure_hotkeys.cpp
index fa9052136..6f7fd4414 100644
--- a/src/yuzu/configuration/configure_hotkeys.cpp
+++ b/src/yuzu/configuration/configure_hotkeys.cpp
@@ -2,10 +2,12 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <QMenu>
 #include <QMessageBox>
 #include <QStandardItemModel>
 #include "core/settings.h"
 #include "ui_configure_hotkeys.h"
+#include "yuzu/configuration/config.h"
 #include "yuzu/configuration/configure_hotkeys.h"
 #include "yuzu/hotkeys.h"
 #include "yuzu/util/sequence_dialog/sequence_dialog.h"
@@ -19,6 +21,9 @@ ConfigureHotkeys::ConfigureHotkeys(QWidget* parent)
     model->setColumnCount(3);
 
     connect(ui->hotkey_list, &QTreeView::doubleClicked, this, &ConfigureHotkeys::Configure);
+    connect(ui->hotkey_list, &QTreeView::customContextMenuRequested, this,
+            &ConfigureHotkeys::PopupContextMenu);
+    ui->hotkey_list->setContextMenuPolicy(Qt::CustomContextMenu);
     ui->hotkey_list->setModel(model);
 
     // TODO(Kloen): Make context configurable as well (hiding the column for now)
@@ -27,6 +32,10 @@ ConfigureHotkeys::ConfigureHotkeys(QWidget* parent)
     ui->hotkey_list->setColumnWidth(0, 200);
     ui->hotkey_list->resizeColumnToContents(1);
 
+    connect(ui->button_restore_defaults, &QPushButton::clicked, this,
+            &ConfigureHotkeys::RestoreDefaults);
+    connect(ui->button_clear_all, &QPushButton::clicked, this, &ConfigureHotkeys::ClearAll);
+
     RetranslateUI();
 }
 
@@ -71,7 +80,6 @@ void ConfigureHotkeys::Configure(QModelIndex index) {
     }
 
     index = index.sibling(index.row(), 1);
-    auto* const model = ui->hotkey_list->model();
     const auto previous_key = model->data(index);
 
     SequenceDialog hotkey_dialog{this};
@@ -81,31 +89,33 @@ void ConfigureHotkeys::Configure(QModelIndex index) {
     if (return_code == QDialog::Rejected || key_sequence.isEmpty()) {
         return;
     }
+    const auto [key_sequence_used, used_action] = IsUsedKey(key_sequence);
 
-    if (IsUsedKey(key_sequence) && key_sequence != QKeySequence(previous_key.toString())) {
-        QMessageBox::warning(this, tr("Conflicting Key Sequence"),
-                             tr("The entered key sequence is already assigned to another hotkey."));
+    if (key_sequence_used && key_sequence != QKeySequence(previous_key.toString())) {
+        QMessageBox::warning(
+            this, tr("Conflicting Key Sequence"),
+            tr("The entered key sequence is already assigned to: %1").arg(used_action));
     } else {
         model->setData(index, key_sequence.toString(QKeySequence::NativeText));
     }
 }
 
-bool ConfigureHotkeys::IsUsedKey(QKeySequence key_sequence) const {
-    for (int r = 0; r < model->rowCount(); r++) {
+std::pair<bool, QString> ConfigureHotkeys::IsUsedKey(QKeySequence key_sequence) const {
+    for (int r = 0; r < model->rowCount(); ++r) {
         const QStandardItem* const parent = model->item(r, 0);
 
-        for (int r2 = 0; r2 < parent->rowCount(); r2++) {
+        for (int r2 = 0; r2 < parent->rowCount(); ++r2) {
             const QStandardItem* const key_seq_item = parent->child(r2, 1);
             const auto key_seq_str = key_seq_item->text();
             const auto key_seq = QKeySequence::fromString(key_seq_str, QKeySequence::NativeText);
 
             if (key_sequence == key_seq) {
-                return true;
+                return std::make_pair(true, parent->child(r2, 0)->text());
             }
         }
     }
 
-    return false;
+    return std::make_pair(false, QString());
 }
 
 void ConfigureHotkeys::ApplyConfiguration(HotkeyRegistry& registry) {
@@ -128,3 +138,55 @@ void ConfigureHotkeys::ApplyConfiguration(HotkeyRegistry& registry) {
 
     registry.SaveHotkeys();
 }
+
+void ConfigureHotkeys::RestoreDefaults() {
+    for (int r = 0; r < model->rowCount(); ++r) {
+        const QStandardItem* parent = model->item(r, 0);
+
+        for (int r2 = 0; r2 < parent->rowCount(); ++r2) {
+            model->item(r, 0)->child(r2, 1)->setText(Config::default_hotkeys[r2].shortcut.first);
+        }
+    }
+}
+
+void ConfigureHotkeys::ClearAll() {
+    for (int r = 0; r < model->rowCount(); ++r) {
+        const QStandardItem* parent = model->item(r, 0);
+
+        for (int r2 = 0; r2 < parent->rowCount(); ++r2) {
+            model->item(r, 0)->child(r2, 1)->setText(tr(""));
+        }
+    }
+}
+
+void ConfigureHotkeys::PopupContextMenu(const QPoint& menu_location) {
+    QModelIndex index = ui->hotkey_list->indexAt(menu_location);
+    if (!index.parent().isValid()) {
+        return;
+    }
+
+    const auto selected = index.sibling(index.row(), 1);
+    QMenu context_menu;
+
+    QAction* restore_default = context_menu.addAction(tr("Restore Default"));
+    QAction* clear = context_menu.addAction(tr("Clear"));
+
+    connect(restore_default, &QAction::triggered, [this, selected] {
+        const QKeySequence& default_key_sequence = QKeySequence::fromString(
+            Config::default_hotkeys[selected.row()].shortcut.first, QKeySequence::NativeText);
+        const auto [key_sequence_used, used_action] = IsUsedKey(default_key_sequence);
+
+        if (key_sequence_used &&
+            default_key_sequence != QKeySequence(model->data(selected).toString())) {
+
+            QMessageBox::warning(
+                this, tr("Conflicting Key Sequence"),
+                tr("The default key sequence is already assigned to: %1").arg(used_action));
+        } else {
+            model->setData(selected, default_key_sequence.toString(QKeySequence::NativeText));
+        }
+    });
+    connect(clear, &QAction::triggered, [this, selected] { model->setData(selected, tr("")); });
+
+    context_menu.exec(ui->hotkey_list->viewport()->mapToGlobal(menu_location));
+}
diff --git a/src/yuzu/configuration/configure_hotkeys.h b/src/yuzu/configuration/configure_hotkeys.h
index 8f8c6173b..a2ec3323e 100644
--- a/src/yuzu/configuration/configure_hotkeys.h
+++ b/src/yuzu/configuration/configure_hotkeys.h
@@ -35,7 +35,11 @@ private:
     void RetranslateUI();
 
     void Configure(QModelIndex index);
-    bool IsUsedKey(QKeySequence key_sequence) const;
+    std::pair<bool, QString> IsUsedKey(QKeySequence key_sequence) const;
+
+    void RestoreDefaults();
+    void ClearAll();
+    void PopupContextMenu(const QPoint& menu_location);
 
     std::unique_ptr<Ui::ConfigureHotkeys> ui;
 
diff --git a/src/yuzu/configuration/configure_hotkeys.ui b/src/yuzu/configuration/configure_hotkeys.ui
index 0d0b70f38..6d9f861e3 100644
--- a/src/yuzu/configuration/configure_hotkeys.ui
+++ b/src/yuzu/configuration/configure_hotkeys.ui
@@ -6,8 +6,8 @@
    <rect>
     <x>0</x>
     <y>0</y>
-    <width>363</width>
-    <height>388</height>
+    <width>439</width>
+    <height>510</height>
    </rect>
   </property>
   <property name="windowTitle">
@@ -15,7 +15,7 @@
   </property>
   <layout class="QVBoxLayout" name="verticalLayout">
    <item>
-    <layout class="QVBoxLayout" name="verticalLayout_2">
+    <layout class="QHBoxLayout" name="horizontalLayout">
      <item>
       <widget class="QLabel" name="label_2">
        <property name="text">
@@ -24,6 +24,37 @@
       </widget>
      </item>
      <item>
+      <spacer name="horizontalSpacer">
+       <property name="orientation">
+        <enum>Qt::Horizontal</enum>
+       </property>
+       <property name="sizeHint" stdset="0">
+        <size>
+         <width>40</width>
+         <height>20</height>
+        </size>
+       </property>
+      </spacer>
+     </item>
+     <item>
+      <widget class="QPushButton" name="button_clear_all">
+       <property name="text">
+        <string>Clear All</string>
+       </property>
+      </widget>
+     </item>
+     <item>
+      <widget class="QPushButton" name="button_restore_defaults">
+       <property name="text">
+        <string>Restore Defaults</string>
+       </property>
+      </widget>
+     </item>
+    </layout>
+   </item>
+   <item>
+    <layout class="QVBoxLayout" name="verticalLayout_2">
+     <item>
       <widget class="QTreeView" name="hotkey_list">
        <property name="editTriggers">
         <set>QAbstractItemView::NoEditTriggers</set>
@@ -39,4 +70,4 @@
  </widget>
  <resources/>
  <connections/>
-</ui>
\ No newline at end of file
+</ui>
diff --git a/src/yuzu/configuration/configure_input_player.cpp b/src/yuzu/configuration/configure_input_player.cpp
index 15ac30f12..e4eb5594b 100644
--- a/src/yuzu/configuration/configure_input_player.cpp
+++ b/src/yuzu/configuration/configure_input_player.cpp
@@ -56,7 +56,6 @@ static void SetAnalogButton(const Common::ParamPackage& input_param,
     if (analog_param.Get("engine", "") != "analog_from_button") {
         analog_param = {
             {"engine", "analog_from_button"},
-            {"modifier_scale", "0.5"},
         };
     }
     analog_param.Set(button_name, input_param.Serialize());
@@ -236,8 +235,10 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i
         widget->setVisible(false);
 
     analog_map_stick = {ui->buttonLStickAnalog, ui->buttonRStickAnalog};
-    analog_map_deadzone = {ui->sliderLStickDeadzone, ui->sliderRStickDeadzone};
-    analog_map_deadzone_label = {ui->labelLStickDeadzone, ui->labelRStickDeadzone};
+    analog_map_deadzone_and_modifier_slider = {ui->sliderLStickDeadzoneAndModifier,
+                                               ui->sliderRStickDeadzoneAndModifier};
+    analog_map_deadzone_and_modifier_slider_label = {ui->labelLStickDeadzoneAndModifier,
+                                                     ui->labelRStickDeadzoneAndModifier};
 
     for (int button_id = 0; button_id < Settings::NativeButton::NumButtons; button_id++) {
         auto* const button = button_map[button_id];
@@ -328,10 +329,18 @@ ConfigureInputPlayer::ConfigureInputPlayer(QWidget* parent, std::size_t player_i
                     InputCommon::Polling::DeviceType::Analog);
             }
         });
-        connect(analog_map_deadzone[analog_id], &QSlider::valueChanged, [=] {
-            const float deadzone = analog_map_deadzone[analog_id]->value() / 100.0f;
-            analog_map_deadzone_label[analog_id]->setText(tr("Deadzone: %1").arg(deadzone));
-            analogs_param[analog_id].Set("deadzone", deadzone);
+
+        connect(analog_map_deadzone_and_modifier_slider[analog_id], &QSlider::valueChanged, [=] {
+            const float slider_value = analog_map_deadzone_and_modifier_slider[analog_id]->value();
+            if (analogs_param[analog_id].Get("engine", "") == "sdl") {
+                analog_map_deadzone_and_modifier_slider_label[analog_id]->setText(
+                    tr("Deadzone: %1%").arg(slider_value));
+                analogs_param[analog_id].Set("deadzone", slider_value / 100.0f);
+            } else {
+                analog_map_deadzone_and_modifier_slider_label[analog_id]->setText(
+                    tr("Modifier Scale: %1%").arg(slider_value));
+                analogs_param[analog_id].Set("modifier_scale", slider_value / 100.0f);
+            }
         });
     }
 
@@ -517,20 +526,31 @@ void ConfigureInputPlayer::UpdateButtonLabels() {
         analog_map_stick[analog_id]->setText(tr("Set Analog Stick"));
 
         auto& param = analogs_param[analog_id];
-        auto* const analog_deadzone_slider = analog_map_deadzone[analog_id];
-        auto* const analog_deadzone_label = analog_map_deadzone_label[analog_id];
-
-        if (param.Has("engine") && param.Get("engine", "") == "sdl") {
-            if (!param.Has("deadzone")) {
-                param.Set("deadzone", 0.1f);
+        auto* const analog_stick_slider = analog_map_deadzone_and_modifier_slider[analog_id];
+        auto* const analog_stick_slider_label =
+            analog_map_deadzone_and_modifier_slider_label[analog_id];
+
+        if (param.Has("engine")) {
+            if (param.Get("engine", "") == "sdl") {
+                if (!param.Has("deadzone")) {
+                    param.Set("deadzone", 0.1f);
+                }
+
+                analog_stick_slider->setValue(static_cast<int>(param.Get("deadzone", 0.1f) * 100));
+                if (analog_stick_slider->value() == 0) {
+                    analog_stick_slider_label->setText(tr("Deadzone: 0%"));
+                }
+            } else {
+                if (!param.Has("modifier_scale")) {
+                    param.Set("modifier_scale", 0.5f);
+                }
+
+                analog_stick_slider->setValue(
+                    static_cast<int>(param.Get("modifier_scale", 0.5f) * 100));
+                if (analog_stick_slider->value() == 0) {
+                    analog_stick_slider_label->setText(tr("Modifier Scale: 0%"));
+                }
             }
-
-            analog_deadzone_slider->setValue(static_cast<int>(param.Get("deadzone", 0.1f) * 100));
-            analog_deadzone_slider->setVisible(true);
-            analog_deadzone_label->setVisible(true);
-        } else {
-            analog_deadzone_slider->setVisible(false);
-            analog_deadzone_label->setVisible(false);
         }
     }
 }
diff --git a/src/yuzu/configuration/configure_input_player.h b/src/yuzu/configuration/configure_input_player.h
index 045704e47..95afa5375 100644
--- a/src/yuzu/configuration/configure_input_player.h
+++ b/src/yuzu/configuration/configure_input_player.h
@@ -97,8 +97,10 @@ private:
     /// Analog inputs are also represented each with a single button, used to configure with an
     /// actual analog stick
     std::array<QPushButton*, Settings::NativeAnalog::NumAnalogs> analog_map_stick;
-    std::array<QSlider*, Settings::NativeAnalog::NumAnalogs> analog_map_deadzone;
-    std::array<QLabel*, Settings::NativeAnalog::NumAnalogs> analog_map_deadzone_label;
+    std::array<QSlider*, Settings::NativeAnalog::NumAnalogs>
+        analog_map_deadzone_and_modifier_slider;
+    std::array<QLabel*, Settings::NativeAnalog::NumAnalogs>
+        analog_map_deadzone_and_modifier_slider_label;
 
     static const std::array<std::string, ANALOG_SUB_BUTTONS_NUM> analog_sub_buttons;
 
diff --git a/src/yuzu/configuration/configure_input_player.ui b/src/yuzu/configuration/configure_input_player.ui
index 4b37746a1..f27a77180 100644
--- a/src/yuzu/configuration/configure_input_player.ui
+++ b/src/yuzu/configuration/configure_input_player.ui
@@ -171,11 +171,11 @@
          </layout>
         </item>
         <item row="4" column="0" colspan="2">
-         <layout class="QVBoxLayout" name="sliderRStickDeadzoneVerticalLayout">
+         <layout class="QVBoxLayout" name="sliderRStickDeadzoneAndModifierVerticalLayout">
           <item>
-           <layout class="QHBoxLayout" name="sliderRStickDeadzoneHorizontalLayout">
+           <layout class="QHBoxLayout" name="sliderRStickDeadzoneAndModifierHorizontalLayout">
             <item>
-             <widget class="QLabel" name="labelRStickDeadzone">
+             <widget class="QLabel" name="labelRStickDeadzoneAndModifier">
               <property name="text">
                <string>Deadzone: 0</string>
               </property>
@@ -187,7 +187,7 @@
            </layout>
           </item>
           <item>
-           <widget class="QSlider" name="sliderRStickDeadzone">
+           <widget class="QSlider" name="sliderRStickDeadzoneAndModifier">
             <property name="orientation">
              <enum>Qt::Horizontal</enum>
             </property>
@@ -784,14 +784,14 @@
          </layout>
         </item>
         <item row="5" column="1" colspan="2">
-         <layout class="QVBoxLayout" name="sliderLStickDeadzoneVerticalLayout">
+         <layout class="QVBoxLayout" name="sliderLStickDeadzoneAndModifierVerticalLayout">
           <property name="sizeConstraint">
            <enum>QLayout::SetDefaultConstraint</enum>
           </property>
           <item>
-           <layout class="QHBoxLayout" name="sliderLStickDeadzoneHorizontalLayout">
+           <layout class="QHBoxLayout" name="sliderLStickDeadzoneAndModifierHorizontalLayout">
             <item>
-             <widget class="QLabel" name="labelLStickDeadzone">
+             <widget class="QLabel" name="labelLStickDeadzoneAndModifier">
               <property name="text">
                <string>Deadzone: 0</string>
               </property>
@@ -803,7 +803,7 @@
            </layout>
           </item>
           <item>
-           <widget class="QSlider" name="sliderLStickDeadzone">
+           <widget class="QSlider" name="sliderLStickDeadzoneAndModifier">
             <property name="orientation">
              <enum>Qt::Horizontal</enum>
             </property>
diff --git a/src/yuzu/game_list_p.h b/src/yuzu/game_list_p.h
index 3e6d5a7cd..0cd0054c8 100644
--- a/src/yuzu/game_list_p.h
+++ b/src/yuzu/game_list_p.h
@@ -126,13 +126,6 @@ public:
 
         return GameListItem::data(role);
     }
-
-    /**
-     * Override to prevent automatic sorting.
-     */
-    bool operator<(const QStandardItem& other) const override {
-        return false;
-    }
 };
 
 class GameListItemCompat : public GameListItem {
@@ -279,6 +272,13 @@ public:
         return static_cast<int>(dir_type);
     }
 
+    /**
+     * Override to prevent automatic sorting between folders and the addDir button.
+     */
+    bool operator<(const QStandardItem& other) const override {
+        return false;
+    }
+
 private:
     GameListItemType dir_type;
 };
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index 80341747f..8476a5a16 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -388,12 +388,14 @@ void Config::ReadValues() {
         static_cast<u16>(sdl2_config->GetInteger("Renderer", "frame_limit", 100));
     Settings::values.use_disk_shader_cache =
         sdl2_config->GetBoolean("Renderer", "use_disk_shader_cache", false);
-    Settings::values.use_accurate_gpu_emulation =
-        sdl2_config->GetBoolean("Renderer", "use_accurate_gpu_emulation", false);
+    const int gpu_accuracy_level = sdl2_config->GetInteger("Renderer", "gpu_accuracy", 0);
+    Settings::values.gpu_accuracy = static_cast<Settings::GPUAccuracy>(gpu_accuracy_level);
     Settings::values.use_asynchronous_gpu_emulation =
         sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", false);
     Settings::values.use_vsync =
         static_cast<u16>(sdl2_config->GetInteger("Renderer", "use_vsync", 1));
+    Settings::values.use_fast_gpu_time =
+        sdl2_config->GetBoolean("Renderer", "use_fast_gpu_time", true);
 
     Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
     Settings::values.bg_green =
diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h
index 171d16fa0..60b1a62fa 100644
--- a/src/yuzu_cmd/default_ini.h
+++ b/src/yuzu_cmd/default_ini.h
@@ -146,9 +146,9 @@ frame_limit =
 # 0 (default): Off, 1 : On
 use_disk_shader_cache =
 
-# Whether to use accurate GPU emulation
-# 0 (default): Off (fast), 1 : On (slow)
-use_accurate_gpu_emulation =
+# Which gpu accuracy level to use
+# 0 (Normal), 1 (High), 2 (Extreme)
+gpu_accuracy =
 
 # Whether to use asynchronous GPU emulation
 # 0 : Off (slow), 1 (default): On (fast)
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
index 19584360c..e5e684206 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2.cpp
@@ -181,9 +181,10 @@ void EmuWindow_SDL2::PollEvents() {
     const u32 current_time = SDL_GetTicks();
     if (current_time > last_time + 2000) {
         const auto results = Core::System::GetInstance().GetAndResetPerfStats();
-        const auto title = fmt::format(
-            "yuzu {} | {}-{} | FPS: {:.0f} ({:.0%})", Common::g_build_fullname,
-            Common::g_scm_branch, Common::g_scm_desc, results.game_fps, results.emulation_speed);
+        const auto title =
+            fmt::format("yuzu {} | {}-{} | FPS: {:.0f} ({:.0f}%)", Common::g_build_fullname,
+                        Common::g_scm_branch, Common::g_scm_desc, results.game_fps,
+                        results.emulation_speed * 100.0);
         SDL_SetWindowTitle(render_window, title.c_str());
         last_time = current_time;
     }
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
index f2990910e..cb8e68a39 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.cpp
@@ -29,6 +29,7 @@ EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(Core::System& system, bool fullscreen)
                          SDL_WINDOW_RESIZABLE | SDL_WINDOW_ALLOW_HIGHDPI);
 
     SDL_SysWMinfo wm;
+    SDL_VERSION(&wm.version);
     if (SDL_GetWindowWMInfo(render_window, &wm) == SDL_FALSE) {
         LOG_CRITICAL(Frontend, "Failed to get information from the window manager");
         std::exit(EXIT_FAILURE);
@@ -70,7 +71,7 @@ EmuWindow_SDL2_VK::EmuWindow_SDL2_VK(Core::System& system, bool fullscreen)
 EmuWindow_SDL2_VK::~EmuWindow_SDL2_VK() = default;
 
 std::unique_ptr<Core::Frontend::GraphicsContext> EmuWindow_SDL2_VK::CreateSharedContext() const {
-    return nullptr;
+    return std::make_unique<DummyContext>();
 }
 
 void EmuWindow_SDL2_VK::Present() {
diff --git a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
index b8021ebea..77a6ca72b 100644
--- a/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
+++ b/src/yuzu_cmd/emu_window/emu_window_sdl2_vk.h
@@ -22,3 +22,5 @@ public:
 
     std::unique_ptr<Core::Frontend::GraphicsContext> CreateSharedContext() const override;
 };
+
+class DummyContext : public Core::Frontend::GraphicsContext {};
diff --git a/src/yuzu_tester/config.cpp b/src/yuzu_tester/config.cpp
index ee2591c8f..3be58b15d 100644
--- a/src/yuzu_tester/config.cpp
+++ b/src/yuzu_tester/config.cpp
@@ -126,10 +126,12 @@ void Config::ReadValues() {
     Settings::values.frame_limit = 100;
     Settings::values.use_disk_shader_cache =
         sdl2_config->GetBoolean("Renderer", "use_disk_shader_cache", false);
-    Settings::values.use_accurate_gpu_emulation =
-        sdl2_config->GetBoolean("Renderer", "use_accurate_gpu_emulation", false);
+    const int gpu_accuracy_level = sdl2_config->GetInteger("Renderer", "gpu_accuracy", 0);
+    Settings::values.gpu_accuracy = static_cast<Settings::GPUAccuracy>(gpu_accuracy_level);
     Settings::values.use_asynchronous_gpu_emulation =
         sdl2_config->GetBoolean("Renderer", "use_asynchronous_gpu_emulation", false);
+    Settings::values.use_fast_gpu_time =
+        sdl2_config->GetBoolean("Renderer", "use_fast_gpu_time", true);
 
     Settings::values.bg_red = static_cast<float>(sdl2_config->GetReal("Renderer", "bg_red", 0.0));
     Settings::values.bg_green =