NvHost: Remake Ctrl Implementation.

SVC: Implement svcInfo:IdleTickCount

CMakeLists.txt

@@ -411,12 +411,13 @@ if (CONAN_REQUIRED_LIBS)
     # Download conan.cmake automatically, you can also just copy the conan.cmake file
     if(NOT EXISTS "${CMAKE_BINARY_DIR}/conan.cmake")
         message(STATUS "Downloading conan.cmake from https://github.com/conan-io/cmake-conan")
-        file(DOWNLOAD "https://github.com/conan-io/cmake-conan/raw/v0.15/conan.cmake"
+        # TODO: Use a tagged release. The latest tagged release does not support VS2022 as of this writing.
+        file(DOWNLOAD "https://raw.githubusercontent.com/conan-io/cmake-conan/43e385830ee35377dbd2dcbe8d5a9e750301ea00/conan.cmake"
                         "${CMAKE_BINARY_DIR}/conan.cmake")
     endif()
     include(${CMAKE_BINARY_DIR}/conan.cmake)
 
-    conan_check(VERSION 1.24.0 REQUIRED)
+    conan_check(VERSION 1.41.0 REQUIRED)
 
     # Manually add iconv to fix a dep conflict between qt and sdl2
     # We don't need to add it through find_package or anything since the other two can find it just fine

src/common/alignment.h

@@ -64,7 +64,7 @@ public:
     using propagate_on_container_copy_assignment = std::true_type;
     using propagate_on_container_move_assignment = std::true_type;
     using propagate_on_container_swap = std::true_type;
-    using is_always_equal = std::true_type;
+    using is_always_equal = std::false_type;
 
     constexpr AlignmentAllocator() noexcept = default;
 
@@ -83,6 +83,11 @@ public:
     struct rebind {
         using other = AlignmentAllocator<T2, Align>;
     };
+
+    template <typename T2, size_t Align2>
+    constexpr bool operator==(const AlignmentAllocator<T2, Align2>&) const noexcept {
+        return std::is_same_v<T, T2> && Align == Align2;
+    }
 };
 
 } // namespace Common

src/common/settings.h

@@ -7,7 +7,6 @@
 #include <algorithm>
 #include <array>
 #include <atomic>
-#include <chrono>
 #include <map>
 #include <optional>
 #include <string>
@@ -487,9 +486,9 @@ struct Values {
     // System
     Setting<std::optional<u32>> rng_seed{std::optional<u32>(), "rng_seed"};
     // Measured in seconds since epoch
-    std::optional<std::chrono::seconds> custom_rtc;
+    std::optional<s64> custom_rtc;
     // Set on game boot, reset on stop. Seconds difference between current time and `custom_rtc`
-    std::chrono::seconds custom_rtc_differential;
+    s64 custom_rtc_differential;
 
     BasicSetting<s32> current_user{0, "current_user"};
     RangedSetting<s32> language_index{1, 0, 17, "language_index"};

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -263,7 +263,7 @@ void ARM_Dynarmic_64::Run() {
 }
 
 void ARM_Dynarmic_64::Step() {
-    cb->InterpreterFallback(jit->GetPC(), 1);
+    jit->Step();
 }
 
 ARM_Dynarmic_64::ARM_Dynarmic_64(System& system_, CPUInterrupts& interrupt_handlers_,

src/core/core.cpp

@@ -140,25 +140,45 @@ struct System::Impl {
           cpu_manager{system}, reporter{system}, applet_manager{system}, time_manager{system} {}
 
     SystemResultStatus Run() {
+        std::unique_lock<std::mutex> lk(suspend_guard);
         status = SystemResultStatus::Success;
 
         kernel.Suspend(false);
         core_timing.SyncPause(false);
         cpu_manager.Pause(false);
+        is_paused = false;
 
         return status;
     }
 
     SystemResultStatus Pause() {
+        std::unique_lock<std::mutex> lk(suspend_guard);
         status = SystemResultStatus::Success;
 
         core_timing.SyncPause(true);
         kernel.Suspend(true);
         cpu_manager.Pause(true);
+        is_paused = true;
 
         return status;
     }
 
+    std::unique_lock<std::mutex> StallCPU() {
+        std::unique_lock<std::mutex> lk(suspend_guard);
+        kernel.Suspend(true);
+        core_timing.SyncPause(true);
+        cpu_manager.Pause(true);
+        return lk;
+    }
+
+    void UnstallCPU() {
+        if (!is_paused) {
+            core_timing.SyncPause(false);
+            kernel.Suspend(false);
+            cpu_manager.Pause(false);
+        }
+    }
+
     SystemResultStatus Init(System& system, Frontend::EmuWindow& emu_window) {
         LOG_DEBUG(Core, "initialized OK");
 
@@ -176,8 +196,9 @@ struct System::Impl {
         cpu_manager.Initialize();
         core_timing.Initialize([&system]() { system.RegisterHostThread(); });
 
-        const auto current_time = std::chrono::duration_cast<std::chrono::seconds>(
-            std::chrono::system_clock::now().time_since_epoch());
+        const auto posix_time = std::chrono::system_clock::now().time_since_epoch();
+        const auto current_time =
+            std::chrono::duration_cast<std::chrono::seconds>(posix_time).count();
         Settings::values.custom_rtc_differential =
             Settings::values.custom_rtc.value_or(current_time) - current_time;
 
@@ -367,6 +388,9 @@ struct System::Impl {
         return perf_stats->GetAndResetStats(core_timing.GetGlobalTimeUs());
     }
 
+    std::mutex suspend_guard;
+    bool is_paused{};
+
     Timing::CoreTiming core_timing;
     Kernel::KernelCore kernel;
     /// RealVfsFilesystem instance
@@ -464,6 +488,14 @@ void System::Shutdown() {
     impl->Shutdown();
 }
 
+std::unique_lock<std::mutex> System::StallCPU() {
+    return impl->StallCPU();
+}
+
+void System::UnstallCPU() {
+    impl->UnstallCPU();
+}
+
 SystemResultStatus System::Load(Frontend::EmuWindow& emu_window, const std::string& filepath,
                                 u64 program_id, std::size_t program_index) {
     return impl->Load(*this, emu_window, filepath, program_id, program_index);

src/core/core.h

@@ -7,6 +7,7 @@
 #include <cstddef>
 #include <functional>
 #include <memory>
+#include <mutex>
 #include <string>
 #include <vector>
 
@@ -160,6 +161,9 @@ public:
     /// Shutdown the emulated system.
     void Shutdown();
 
+    std::unique_lock<std::mutex> StallCPU();
+    void UnstallCPU();
+
     /**
      * Load an executable application.
      * @param emu_window Reference to the host-system window used for video output and keyboard

src/core/hle/kernel/k_page_table.cpp

@@ -363,6 +363,8 @@ ResultCode KPageTable::UnmapProcessCodeMemory(VAddr dst_addr, VAddr src_addr, st
     block_manager->Update(src_addr, num_pages, KMemoryState::Normal,
                           KMemoryPermission::ReadAndWrite);
 
+    system.InvalidateCpuInstructionCacheRange(dst_addr, size);
+
     return ResultSuccess;
 }
 

src/core/hle/kernel/k_scheduler.h

@@ -49,6 +49,11 @@ public:
     /// Gets the current running thread
     [[nodiscard]] KThread* GetCurrentThread() const;
 
+    /// Gets the idle thread
+    [[nodiscard]] KThread* GetIdleThread() const {
+        return idle_thread;
+    }
+
     /// Returns true if the scheduler is idle
     [[nodiscard]] bool IsIdle() const {
         return GetCurrentThread() == idle_thread;

src/core/hle/kernel/svc.cpp

@@ -886,7 +886,24 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, Handle
         *result = out_ticks;
         return ResultSuccess;
     }
+    case GetInfoType::IdleTickCount: {
+        if (handle == 0) {
+            LOG_ERROR(Kernel_SVC, "Thread handle does not exist, handle=0x{:08X}",
+                      static_cast<Handle>(handle));
+            return ResultInvalidHandle;
+        }
 
+        if (info_sub_id != 0xFFFFFFFFFFFFFFFF && info_sub_id != system.CurrentCoreIndex()) {
+            LOG_ERROR(Kernel_SVC, "Core is not the current core, got {}", info_sub_id);
+            return ResultInvalidCombination;
+        }
+
+        const auto& scheduler = *system.Kernel().CurrentScheduler();
+        const auto* const idle_thread = scheduler.GetIdleThread();
+
+        *result = idle_thread->GetCpuTime();
+        return ResultSuccess;
+    }
     default:
         LOG_ERROR(Kernel_SVC, "Unimplemented svcGetInfo id=0x{:016X}", info_id);
         return ResultInvalidEnumValue;

src/core/hle/service/nvdrv/devices/nvhost_ctrl.cpp

@@ -1,4 +1,5 @@
-// Copyright 2018 yuzu emulator team
+// Copyright 2021 yuzu emulator team
+// Copyright 2021 Skyline Team and Contributors (https://github.com/skyline-emu/)
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
@@ -7,6 +8,7 @@
 
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/scope_exit.h"
 #include "core/core.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_writable_event.h"
@@ -72,85 +74,127 @@ NvResult nvhost_ctrl::NvOsGetConfigU32(const std::vector<u8>& input, std::vector
 }
 
 NvResult nvhost_ctrl::IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output,
-                                       bool is_async) {
+                                       bool is_allocation) {
     IocCtrlEventWaitParams params{};
     std::memcpy(&params, input.data(), sizeof(params));
-    LOG_DEBUG(Service_NVDRV, "syncpt_id={}, threshold={}, timeout={}, is_async={}",
-              params.syncpt_id, params.threshold, params.timeout, is_async);
+    LOG_DEBUG(Service_NVDRV, "syncpt_id={}, threshold={}, timeout={}, is_allocation={}",
+              params.fence.id, params.fence.value, params.timeout, is_allocation);
 
-    if (params.syncpt_id >= MaxSyncPoints) {
+    bool must_unmark_fail = !is_allocation;
+    const u32 event_id = params.value.raw;
+    SCOPE_EXIT({
+        std::memcpy(output.data(), &params, sizeof(params));
+        if (must_unmark_fail) {
+            events_interface.fails[event_id] = 0;
+        }
+    });
+
+    const u32 fence_id = static_cast<u32>(params.fence.id);
+
+    if (fence_id >= MaxSyncPoints) {
         return NvResult::BadParameter;
     }
 
-    u32 event_id = params.value & 0x00FF;
-
-    if (event_id >= MaxNvEvents) {
-        std::memcpy(output.data(), &params, sizeof(params));
-        return NvResult::BadParameter;
-    }
-
-    if (syncpoint_manager.IsSyncpointExpired(params.syncpt_id, params.threshold)) {
-        params.value = syncpoint_manager.GetSyncpointMin(params.syncpt_id);
-        std::memcpy(output.data(), &params, sizeof(params));
+    if (params.fence.value == 0) {
         return NvResult::Success;
     }
 
-    if (const auto new_value = syncpoint_manager.RefreshSyncpoint(params.syncpt_id);
-        syncpoint_manager.IsSyncpointExpired(params.syncpt_id, params.threshold)) {
-        params.value = new_value;
-        std::memcpy(output.data(), &params, sizeof(params));
+    if (syncpoint_manager.IsSyncpointExpired(fence_id, params.fence.value)) {
+        params.value.raw = syncpoint_manager.GetSyncpointMin(fence_id);
+        return NvResult::Success;
+    }
+
+    if (const auto new_value = syncpoint_manager.RefreshSyncpoint(fence_id);
+        syncpoint_manager.IsSyncpointExpired(fence_id, params.fence.value)) {
+        params.value.raw = new_value;
         return NvResult::Success;
     }
 
-    auto& event = events_interface.events[event_id];
     auto& gpu = system.GPU();
+    const u32 target_value = params.fence.value;
 
-    // This is mostly to take into account unimplemented features. As synced
-    // gpu is always synced.
-    if (!gpu.IsAsync()) {
-        event.event->GetWritableEvent().Signal();
-        return NvResult::Success;
-    }
-    const u32 current_syncpoint_value = event.fence.value;
-    const s32 diff = current_syncpoint_value - params.threshold;
-    if (diff >= 0) {
-        event.event->GetWritableEvent().Signal();
-        params.value = current_syncpoint_value;
-        std::memcpy(output.data(), &params, sizeof(params));
-        return NvResult::Success;
-    }
-    const u32 target_value = current_syncpoint_value - diff;
-
-    if (!is_async) {
-        params.value = 0;
-    }
+    const auto check_failing = [&]() {
+        if (!is_allocation && events_interface.fails[event_id] > 1) {
+            {
+                auto lk = system.StallCPU();
+                gpu.WaitFence(fence_id, target_value);
+                system.UnstallCPU();
+            }
+            return true;
+        }
+        return false;
+    };
 
     if (params.timeout == 0) {
-        std::memcpy(output.data(), &params, sizeof(params));
+        if (check_failing()) {
+            return NvResult::Success;
+        }
         return NvResult::Timeout;
     }
 
-    EventState status = events_interface.status[event_id];
-    const bool bad_parameter = status != EventState::Free && status != EventState::Registered;
-    if (bad_parameter) {
-        std::memcpy(output.data(), &params, sizeof(params));
+    auto lock = events_interface.Lock();
+
+    u32 slot = [&]() {
+        if (is_allocation) {
+            params.value.raw = 0;
+            return events_interface.FindFreeEvent(fence_id);
+        } else {
+            return params.value.raw;
+        }
+    }();
+
+    if (slot >= MaxNvEvents) {
         return NvResult::BadParameter;
     }
-    events_interface.SetEventStatus(event_id, EventState::Waiting);
-    events_interface.assigned_syncpt[event_id] = params.syncpt_id;
-    events_interface.assigned_value[event_id] = target_value;
-    if (is_async) {
-        params.value = params.syncpt_id << 4;
-    } else {
-        params.value = ((params.syncpt_id & 0xfff) << 16) | 0x10000000;
+
+    auto* event = events_interface.events[slot].event;
+
+    if (!event) {
+        return NvResult::BadParameter;
     }
-    params.value |= event_id;
-    event.event->GetWritableEvent().Clear();
-    gpu.RegisterSyncptInterrupt(params.syncpt_id, target_value);
-    std::memcpy(output.data(), &params, sizeof(params));
+
+    if (events_interface.IsBeingUsed(slot)) {
+        return NvResult::BadParameter;
+    }
+
+    if (check_failing()) {
+        return NvResult::Success;
+    }
+
+    params.value.raw = 0;
+
+    events_interface.status[slot].store(EventState::Waiting, std::memory_order_relaxed);
+    events_interface.assigned_syncpt[slot] = fence_id;
+    events_interface.assigned_value[slot] = target_value;
+    if (is_allocation) {
+        params.value.syncpoint_id_for_allocation.Assign(static_cast<u16>(fence_id));
+        params.value.event_allocated.Assign(1);
+    } else {
+        params.value.syncpoint_id.Assign(fence_id);
+    }
+    params.value.raw |= slot;
+
+    gpu.RegisterSyncptInterrupt(fence_id, target_value);
     return NvResult::Timeout;
 }
 
+NvResult nvhost_ctrl::FreeEvent(u32 slot) {
+    if (slot >= MaxNvEvents) {
+        return NvResult::BadParameter;
+    }
+
+    if (!events_interface.registered[slot]) {
+        return NvResult::Success;
+    }
+
+    if (events_interface.IsBeingUsed(slot)) {
+        return NvResult::Busy;
+    }
+
+    events_interface.Free(slot);
+    return NvResult::Success;
+}
+
 NvResult nvhost_ctrl::IocCtrlEventRegister(const std::vector<u8>& input, std::vector<u8>& output) {
     IocCtrlEventRegisterParams params{};
     std::memcpy(&params, input.data(), sizeof(params));
@@ -159,16 +203,16 @@ NvResult nvhost_ctrl::IocCtrlEventRegister(const std::vector<u8>& input, std::ve
     if (event_id >= MaxNvEvents) {
         return NvResult::BadParameter;
     }
+
+    auto lock = events_interface.Lock();
+
     if (events_interface.registered[event_id]) {
-        const auto event_state = events_interface.status[event_id];
-        if (event_state != EventState::Free) {
-            LOG_WARNING(Service_NVDRV, "Event already registered! Unregistering previous event");
-            events_interface.UnregisterEvent(event_id);
-        } else {
-            return NvResult::BadParameter;
+        const auto result = FreeEvent(event_id);
+        if (result != NvResult::Success) {
+            return result;
         }
     }
-    events_interface.RegisterEvent(event_id);
+    events_interface.Create(event_id);
     return NvResult::Success;
 }
 
@@ -178,31 +222,33 @@ NvResult nvhost_ctrl::IocCtrlEventUnregister(const std::vector<u8>& input,
     std::memcpy(&params, input.data(), sizeof(params));
     const u32 event_id = params.user_event_id & 0x00FF;
     LOG_DEBUG(Service_NVDRV, " called, user_event_id: {:X}", event_id);
-    if (event_id >= MaxNvEvents) {
-        return NvResult::BadParameter;
-    }
-    if (!events_interface.registered[event_id]) {
-        return NvResult::BadParameter;
-    }
-    events_interface.UnregisterEvent(event_id);
-    return NvResult::Success;
+
+    auto lock = events_interface.Lock();
+    return FreeEvent(event_id);
 }
 
 NvResult nvhost_ctrl::IocCtrlClearEventWait(const std::vector<u8>& input, std::vector<u8>& output) {
-    IocCtrlEventSignalParams params{};
+    IocCtrlEventClearParams params{};
     std::memcpy(&params, input.data(), sizeof(params));
 
-    u32 event_id = params.event_id & 0x00FF;
-    LOG_WARNING(Service_NVDRV, "cleared event wait on, event_id: {:X}", event_id);
+    u32 event_id = params.event_id.slot;
+    LOG_DEBUG(Service_NVDRV, "cleared event wait on, event_id: {:X}", event_id);
 
     if (event_id >= MaxNvEvents) {
         return NvResult::BadParameter;
     }
-    if (events_interface.status[event_id] == EventState::Waiting) {
-        events_interface.LiberateEvent(event_id);
-    }
 
-    syncpoint_manager.RefreshSyncpoint(events_interface.events[event_id].fence.id);
+    auto lock = events_interface.Lock();
+
+    if (events_interface.status[event_id].exchange(
+            EventState::Cancelling, std::memory_order_acq_rel) == EventState::Waiting) {
+        system.GPU().CancelSyncptInterrupt(events_interface.assigned_syncpt[event_id],
+                                           events_interface.assigned_value[event_id]);
+        syncpoint_manager.RefreshSyncpoint(events_interface.events[event_id].fence.id);
+    }
+    events_interface.fails[event_id]++;
+    events_interface.status[event_id].store(EventState::Cancelled, std::memory_order_release);
+    events_interface.events[event_id].event->GetWritableEvent().Clear();
 
     return NvResult::Success;
 }

src/core/hle/service/nvdrv/devices/nvhost_ctrl.h

@@ -6,6 +6,7 @@
 
 #include <array>
 #include <vector>
+#include "common/bit_field.h"
 #include "common/common_types.h"
 #include "core/hle/service/nvdrv/devices/nvdevice.h"
 #include "core/hle/service/nvdrv/nvdrv.h"
@@ -28,6 +29,24 @@ public:
     void OnOpen(DeviceFD fd) override;
     void OnClose(DeviceFD fd) override;
 
+    union SyncpointEventValue {
+        u32 raw;
+
+        union {
+            BitField<0, 4, u32> partial_slot;
+            BitField<4, 28, u32> syncpoint_id;
+        };
+
+        struct {
+            u16 slot;
+            union {
+                BitField<0, 12, u16> syncpoint_id_for_allocation;
+                BitField<12, 1, u16> event_allocated;
+            };
+        };
+    };
+    static_assert(sizeof(SyncpointEventValue) == sizeof(u32));
+
 private:
     struct IocSyncptReadParams {
         u32_le id{};
@@ -84,27 +103,18 @@ private:
     };
     static_assert(sizeof(IocGetConfigParams) == 387, "IocGetConfigParams is incorrect size");
 
-    struct IocCtrlEventSignalParams {
-        u32_le event_id{};
+    struct IocCtrlEventClearParams {
+        SyncpointEventValue event_id{};
     };
-    static_assert(sizeof(IocCtrlEventSignalParams) == 4,
-                  "IocCtrlEventSignalParams is incorrect size");
+    static_assert(sizeof(IocCtrlEventClearParams) == 4,
+                  "IocCtrlEventClearParams is incorrect size");
 
     struct IocCtrlEventWaitParams {
-        u32_le syncpt_id{};
-        u32_le threshold{};
-        s32_le timeout{};
-        u32_le value{};
-    };
-    static_assert(sizeof(IocCtrlEventWaitParams) == 16, "IocCtrlEventWaitParams is incorrect size");
-
-    struct IocCtrlEventWaitAsyncParams {
-        u32_le syncpt_id{};
-        u32_le threshold{};
+        Fence fence{};
         u32_le timeout{};
-        u32_le value{};
+        SyncpointEventValue value{};
     };
-    static_assert(sizeof(IocCtrlEventWaitAsyncParams) == 16,
+    static_assert(sizeof(IocCtrlEventWaitParams) == 16,
                   "IocCtrlEventWaitAsyncParams is incorrect size");
 
     struct IocCtrlEventRegisterParams {
@@ -125,11 +135,14 @@ private:
     static_assert(sizeof(IocCtrlEventKill) == 8, "IocCtrlEventKill is incorrect size");
 
     NvResult NvOsGetConfigU32(const std::vector<u8>& input, std::vector<u8>& output);
-    NvResult IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output, bool is_async);
+    NvResult IocCtrlEventWait(const std::vector<u8>& input, std::vector<u8>& output,
+                              bool is_allocation);
     NvResult IocCtrlEventRegister(const std::vector<u8>& input, std::vector<u8>& output);
     NvResult IocCtrlEventUnregister(const std::vector<u8>& input, std::vector<u8>& output);
     NvResult IocCtrlClearEventWait(const std::vector<u8>& input, std::vector<u8>& output);
 
+    NvResult FreeEvent(u32 slot);
+
     EventInterface& events_interface;
     SyncpointManager& syncpoint_manager;
 };

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp

@@ -69,8 +69,7 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
     std::vector<Reloc> relocs(params.relocation_count);
     std::vector<u32> reloc_shifts(params.relocation_count);
     std::vector<SyncptIncr> syncpt_increments(params.syncpoint_count);
-    std::vector<SyncptIncr> wait_checks(params.syncpoint_count);
-    std::vector<Fence> fences(params.fence_count);
+    std::vector<u32> fence_thresholds(params.fence_count);
 
     // Slice input into their respective buffers
     std::size_t offset = sizeof(IoctlSubmit);
@@ -78,15 +77,13 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
     offset += SliceVectors(input, relocs, params.relocation_count, offset);
     offset += SliceVectors(input, reloc_shifts, params.relocation_count, offset);
     offset += SliceVectors(input, syncpt_increments, params.syncpoint_count, offset);
-    offset += SliceVectors(input, wait_checks, params.syncpoint_count, offset);
-    offset += SliceVectors(input, fences, params.fence_count, offset);
+    offset += SliceVectors(input, fence_thresholds, params.fence_count, offset);
 
     auto& gpu = system.GPU();
     if (gpu.UseNvdec()) {
         for (std::size_t i = 0; i < syncpt_increments.size(); i++) {
             const SyncptIncr& syncpt_incr = syncpt_increments[i];
-            fences[i].id = syncpt_incr.id;
-            fences[i].value =
+            fence_thresholds[i] =
                 syncpoint_manager.IncreaseSyncpoint(syncpt_incr.id, syncpt_incr.increments);
         }
     }
@@ -98,11 +95,6 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
                                   cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(cmdlist);
     }
-    if (gpu.UseNvdec()) {
-        fences[0].value = syncpoint_manager.IncreaseSyncpoint(fences[0].id, 1);
-        Tegra::ChCommandHeaderList cmdlist{{(4 << 28) | fences[0].id}};
-        gpu.PushCommandBuffer(cmdlist);
-    }
     std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
     // Some games expect command_buffers to be written back
     offset = sizeof(IoctlSubmit);
@@ -110,8 +102,7 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
     offset += WriteVectors(output, relocs, offset);
     offset += WriteVectors(output, reloc_shifts, offset);
     offset += WriteVectors(output, syncpt_increments, offset);
-    offset += WriteVectors(output, wait_checks, offset);
-    offset += WriteVectors(output, fences, offset);
+    offset += WriteVectors(output, fence_thresholds, offset);
 
     return NvResult::Success;
 }

src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.h

@@ -56,19 +56,16 @@ protected:
         s32 target{};
         s32 target_offset{};
     };
-    static_assert(sizeof(Reloc) == 0x10, "CommandBuffer has incorrect size");
+    static_assert(sizeof(Reloc) == 0x10, "Reloc has incorrect size");
 
     struct SyncptIncr {
         u32 id{};
         u32 increments{};
+        u32 unk0{};
+        u32 unk1{};
+        u32 unk2{};
     };
-    static_assert(sizeof(SyncptIncr) == 0x8, "CommandBuffer has incorrect size");
-
-    struct Fence {
-        u32 id{};
-        u32 value{};
-    };
-    static_assert(sizeof(Fence) == 0x8, "CommandBuffer has incorrect size");
+    static_assert(sizeof(SyncptIncr) == 0x14, "SyncptIncr has incorrect size");
 
     struct IoctlGetSyncpoint {
         // Input

src/core/hle/service/nvdrv/nvdata.h

@@ -1,3 +1,8 @@
+// Copyright 2021 yuzu emulator team
+// Copyright 2021 Skyline Team and Contributors (https://github.com/skyline-emu/)
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
 #pragma once
 
 #include <array>
@@ -81,11 +86,15 @@ enum class NvResult : u32 {
     ModuleNotPresent = 0xA000E,
 };
 
+// obtained from
+// https://github.com/skyline-emu/skyline/blob/nvdec-dev/app/src/main/cpp/skyline/services/nvdrv/devices/nvhost/ctrl.h#L47
 enum class EventState {
-    Free = 0,
-    Registered = 1,
-    Waiting = 2,
-    Busy = 3,
+    Available = 0,
+    Waiting = 1,
+    Cancelling = 2,
+    Signalling = 3,
+    Signalled = 4,
+    Cancelled = 5,
 };
 
 union Ioctl {

src/core/hle/service/nvdrv/nvdrv.cpp

@@ -1,7 +1,9 @@
-// Copyright 2018 yuzu emulator team
+// Copyright 2021 yuzu emulator team
+// Copyright 2021 Skyline Team and Contributors (https://github.com/skyline-emu/)
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <bit>
 #include <utility>
 
 #include <fmt/format.h>
@@ -28,6 +30,64 @@
 
 namespace Service::Nvidia {
 
+std::unique_lock<std::mutex> EventInterface::Lock() {
+    return std::unique_lock<std::mutex>(events_mutex);
+}
+
+void EventInterface::Signal(u32 event_id) {
+    if (status[event_id].exchange(EventState::Signalling, std::memory_order_acq_rel) ==
+        EventState::Waiting) {
+        events[event_id].event->GetWritableEvent().Signal();
+    }
+    status[event_id].store(EventState::Signalled, std::memory_order_release);
+}
+
+void EventInterface::Create(u32 event_id) {
+    events[event_id].event =
+        module.service_context.CreateEvent(fmt::format("NVDRV::NvEvent_{}", event_id));
+    status[event_id] = EventState::Available;
+    registered[event_id] = true;
+    const u64 mask = 1ULL << event_id;
+    events_mask |= mask;
+}
+
+void EventInterface::Free(u32 event_id) {
+    module.service_context.CloseEvent(events[event_id].event);
+    events[event_id].event = nullptr;
+    status[event_id] = EventState::Available;
+    registered[event_id] = false;
+    const u64 mask = ~(1ULL << event_id);
+    events_mask &= mask;
+}
+
+u32 EventInterface::FindFreeEvent(u32 syncpoint_id) {
+    u32 slot{MaxNvEvents};
+    u32 free_slot{MaxNvEvents};
+    for (u32 i = 0; i < MaxNvEvents; i++) {
+        if (events[i].event) {
+            if (!IsBeingUsed(i)) {
+                slot = i;
+                if (assigned_syncpt[i] == syncpoint_id) {
+                    return slot;
+                }
+            }
+        } else if (free_slot == MaxNvEvents) {
+            free_slot = i;
+        }
+    }
+    if (free_slot < MaxNvEvents) {
+        Create(free_slot);
+        return free_slot;
+    }
+
+    if (slot < MaxNvEvents) {
+        return slot;
+    }
+
+    LOG_CRITICAL(Service_NVDRV, "Failed to allocate an event");
+    return 0;
+}
+
 void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger& nvflinger,
                        Core::System& system) {
     auto module_ = std::make_shared<Module>(system);
@@ -40,11 +100,11 @@ void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger
 }
 
 Module::Module(Core::System& system)
-    : syncpoint_manager{system.GPU()}, service_context{system, "nvdrv"} {
+    : syncpoint_manager{system.GPU()}, events_interface{*this}, service_context{system, "nvdrv"} {
+    events_interface.events_mask = 0;
     for (u32 i = 0; i < MaxNvEvents; i++) {
-        events_interface.events[i].event =
-            service_context.CreateEvent(fmt::format("NVDRV::NvEvent_{}", i));
-        events_interface.status[i] = EventState::Free;
+        events_interface.status[i] = EventState::Available;
+        events_interface.events[i].event = nullptr;
         events_interface.registered[i] = false;
     }
     auto nvmap_dev = std::make_shared<Devices::nvmap>(system);
@@ -64,8 +124,11 @@ Module::Module(Core::System& system)
 }
 
 Module::~Module() {
+    auto lock = events_interface.Lock();
     for (u32 i = 0; i < MaxNvEvents; i++) {
-        service_context.CloseEvent(events_interface.events[i].event);
+        if (events_interface.registered[i]) {
+            service_context.CloseEvent(events_interface.events[i].event);
+        }
     }
 }
 
@@ -171,21 +234,35 @@ NvResult Module::Close(DeviceFD fd) {
 }
 
 void Module::SignalSyncpt(const u32 syncpoint_id, const u32 value) {
-    for (u32 i = 0; i < MaxNvEvents; i++) {
+    const u32 max = MaxNvEvents - std::countl_zero(events_interface.events_mask);
+    const u32 min = std::countr_zero(events_interface.events_mask);
+    for (u32 i = min; i < max; i++) {
         if (events_interface.assigned_syncpt[i] == syncpoint_id &&
             events_interface.assigned_value[i] == value) {
-            events_interface.LiberateEvent(i);
-            events_interface.events[i].event->GetWritableEvent().Signal();
+            events_interface.Signal(i);
         }
     }
 }
 
-Kernel::KReadableEvent& Module::GetEvent(const u32 event_id) {
-    return events_interface.events[event_id].event->GetReadableEvent();
-}
+Kernel::KEvent* Module::GetEvent(u32 event_id) {
+    const auto event = Devices::nvhost_ctrl::SyncpointEventValue{.raw = event_id};
 
-Kernel::KWritableEvent& Module::GetEventWriteable(const u32 event_id) {
-    return events_interface.events[event_id].event->GetWritableEvent();
+    const bool allocated = event.event_allocated.Value() != 0;
+    const u32 slot{allocated ? event.partial_slot.Value() : static_cast<u32>(event.slot)};
+    if (slot >= MaxNvEvents) {
+        return nullptr;
+    }
+
+    const u32 syncpoint_id{allocated ? event.syncpoint_id_for_allocation.Value()
+                                     : event.syncpoint_id.Value()};
+
+    auto lock = events_interface.Lock();
+
+    if (events_interface.registered[slot] &&
+        events_interface.assigned_syncpt[slot] == syncpoint_id) {
+        return events_interface.events[slot].event;
+    }
+    return nullptr;
 }
 
 } // namespace Service::Nvidia

src/core/hle/service/nvdrv/nvdrv.h

@@ -1,4 +1,5 @@
 // Copyright 2018 yuzu emulator team
+// Copyright 2021 Skyline Team and Contributors (https://github.com/skyline-emu/)
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
@@ -40,15 +41,22 @@ struct NvEvent {
     Fence fence{};
 };
 
-struct EventInterface {
-    // Mask representing currently busy events
+class Module;
+
+class EventInterface {
+public:
+    EventInterface(Module& module_) : module{module_} {}
+
+    // Mask representing registered events
     u64 events_mask{};
     // Each kernel event associated to an NV event
-    std::array<NvEvent, MaxNvEvents> events;
+    std::array<NvEvent, MaxNvEvents> events{};
     // The status of the current NVEvent
-    std::array<EventState, MaxNvEvents> status{};
+    std::array<std::atomic<EventState>, MaxNvEvents> status{};
     // Tells if an NVEvent is registered or not
     std::array<bool, MaxNvEvents> registered{};
+    // Tells the NVEvent that it has failed.
+    std::array<u32, MaxNvEvents> fails{};
     // When an NVEvent is waiting on GPU interrupt, this is the sync_point
     // associated with it.
     std::array<u32, MaxNvEvents> assigned_syncpt{};
@@ -57,50 +65,26 @@ struct EventInterface {
     std::array<u32, MaxNvEvents> assigned_value{};
     // Constant to denote an unasigned syncpoint.
     static constexpr u32 unassigned_syncpt = 0xFFFFFFFF;
-    std::optional<u32> GetFreeEvent() const {
-        u64 mask = events_mask;
-        for (u32 i = 0; i < MaxNvEvents; i++) {
-            const bool is_free = (mask & 0x1) == 0;
-            if (is_free) {
-                if (status[i] == EventState::Registered || status[i] == EventState::Free) {
-                    return {i};
-                }
-            }
-            mask = mask >> 1;
-        }
-        return std::nullopt;
-    }
-    void SetEventStatus(const u32 event_id, EventState new_status) {
-        EventState old_status = status[event_id];
-        if (old_status == new_status) {
-            return;
-        }
-        status[event_id] = new_status;
-        if (new_status == EventState::Registered) {
-            registered[event_id] = true;
-        }
-        if (new_status == EventState::Waiting || new_status == EventState::Busy) {
-            events_mask |= (1ULL << event_id);
-        }
-    }
-    void RegisterEvent(const u32 event_id) {
-        registered[event_id] = true;
-        if (status[event_id] == EventState::Free) {
-            status[event_id] = EventState::Registered;
-        }
-    }
-    void UnregisterEvent(const u32 event_id) {
-        registered[event_id] = false;
-        if (status[event_id] == EventState::Registered) {
-            status[event_id] = EventState::Free;
-        }
-    }
-    void LiberateEvent(const u32 event_id) {
-        status[event_id] = registered[event_id] ? EventState::Registered : EventState::Free;
-        events_mask &= ~(1ULL << event_id);
-        assigned_syncpt[event_id] = unassigned_syncpt;
-        assigned_value[event_id] = 0;
+
+    bool IsBeingUsed(u32 event_id) {
+        const auto current_status = status[event_id].load(std::memory_order_acquire);
+        return current_status == EventState::Waiting || current_status == EventState::Cancelling ||
+               current_status == EventState::Signalling;
     }
+
+    std::unique_lock<std::mutex> Lock();
+
+    void Signal(u32 event_id);
+
+    void Create(u32 event_id);
+
+    void Free(u32 event_id);
+
+    u32 FindFreeEvent(u32 syncpoint_id);
+
+private:
+    std::mutex events_mutex;
+    Module& module;
 };
 
 class Module final {
@@ -137,11 +121,11 @@ public:
 
     void SignalSyncpt(const u32 syncpoint_id, const u32 value);
 
-    Kernel::KReadableEvent& GetEvent(u32 event_id);
-
-    Kernel::KWritableEvent& GetEventWriteable(u32 event_id);
+    Kernel::KEvent* GetEvent(u32 event_id);
 
 private:
+    friend class EventInterface;
+
     /// Manages syncpoints on the host
     SyncpointManager syncpoint_manager;
 
@@ -157,6 +141,9 @@ private:
     EventInterface events_interface;
 
     KernelHelpers::ServiceContext service_context;
+
+    void CreateEvent(u32 event_id);
+    void FreeEvent(u32 event_id);
 };
 
 /// Registers all NVDRV services with the specified service manager.

src/core/hle/service/nvdrv/nvdrv_interface.cpp

@@ -6,6 +6,7 @@
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/hle/ipc_helpers.h"
+#include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_readable_event.h"
 #include "core/hle/kernel/k_thread.h"
 #include "core/hle/kernel/k_writable_event.h"
@@ -168,7 +169,7 @@ void NVDRV::Initialize(Kernel::HLERequestContext& ctx) {
 void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto fd = rp.Pop<DeviceFD>();
-    const auto event_id = rp.Pop<u32>() & 0x00FF;
+    const auto event_id = rp.Pop<u32>();
     LOG_WARNING(Service_NVDRV, "(STUBBED) called, fd={:X}, event_id={:X}", fd, event_id);
 
     if (!is_initialized) {
@@ -184,12 +185,14 @@ void NVDRV::QueryEvent(Kernel::HLERequestContext& ctx) {
         return;
     }
 
-    if (event_id < MaxNvEvents) {
+    auto* event = nvdrv->GetEvent(event_id);
+
+    if (event) {
         IPC::ResponseBuilder rb{ctx, 3, 1};
         rb.Push(ResultSuccess);
-        auto& event = nvdrv->GetEvent(event_id);
-        event.Clear();
-        rb.PushCopyObjects(event);
+        auto& readable_event = event->GetReadableEvent();
+        readable_event.Clear();
+        rb.PushCopyObjects(readable_event);
         rb.PushEnum(NvResult::Success);
     } else {
         IPC::ResponseBuilder rb{ctx, 3};

src/core/hle/service/time/time_manager.cpp

@@ -13,18 +13,19 @@
 #include "core/hle/service/time/time_manager.h"
 
 namespace Service::Time {
-
+namespace {
 constexpr Clock::TimeSpanType standard_network_clock_accuracy{0x0009356907420000ULL};
 
-static std::chrono::seconds GetSecondsSinceEpoch() {
-    return std::chrono::duration_cast<std::chrono::seconds>(
-               std::chrono::system_clock::now().time_since_epoch()) +
+s64 GetSecondsSinceEpoch() {
+    const auto time_since_epoch = std::chrono::system_clock::now().time_since_epoch();
+    return std::chrono::duration_cast<std::chrono::seconds>(time_since_epoch).count() +
            Settings::values.custom_rtc_differential;
 }
 
-static s64 GetExternalRtcValue() {
-    return GetSecondsSinceEpoch().count() + TimeManager::GetExternalTimeZoneOffset();
+s64 GetExternalRtcValue() {
+    return GetSecondsSinceEpoch() + TimeManager::GetExternalTimeZoneOffset();
 }
+} // Anonymous namespace
 
 struct TimeManager::Impl final {
     explicit Impl(Core::System& system)

src/input_common/udp/client.cpp

@@ -471,46 +471,42 @@ CalibrationConfigurationJob::CalibrationConfigurationJob(
     std::function<void(u16, u16, u16, u16)> data_callback) {
 
     std::thread([=, this] {
-        constexpr u16 CALIBRATION_THRESHOLD = 100;
-
-        u16 min_x{UINT16_MAX};
-        u16 min_y{UINT16_MAX};
-        u16 max_x{};
-        u16 max_y{};
-
         Status current_status{Status::Initialized};
-        SocketCallback callback{[](Response::Version) {}, [](Response::PortInfo) {},
-                                [&](Response::PadData data) {
-                                    if (current_status == Status::Initialized) {
-                                        // Receiving data means the communication is ready now
-                                        current_status = Status::Ready;
-                                        status_callback(current_status);
-                                    }
-                                    if (data.touch[0].is_active == 0) {
-                                        return;
-                                    }
-                                    LOG_DEBUG(Input, "Current touch: {} {}", data.touch[0].x,
-                                              data.touch[0].y);
-                                    min_x = std::min(min_x, static_cast<u16>(data.touch[0].x));
-                                    min_y = std::min(min_y, static_cast<u16>(data.touch[0].y));
-                                    if (current_status == Status::Ready) {
-                                        // First touch - min data (min_x/min_y)
-                                        current_status = Status::Stage1Completed;
-                                        status_callback(current_status);
-                                    }
-                                    if (data.touch[0].x - min_x > CALIBRATION_THRESHOLD &&
-                                        data.touch[0].y - min_y > CALIBRATION_THRESHOLD) {
-                                        // Set the current position as max value and finishes
-                                        // configuration
-                                        max_x = data.touch[0].x;
-                                        max_y = data.touch[0].y;
-                                        current_status = Status::Completed;
-                                        data_callback(min_x, min_y, max_x, max_y);
-                                        status_callback(current_status);
+        SocketCallback callback{
+            [](Response::Version) {}, [](Response::PortInfo) {},
+            [&](Response::PadData data) {
+                static constexpr u16 CALIBRATION_THRESHOLD = 100;
+                static constexpr u16 MAX_VALUE = UINT16_MAX;
 
-                                        complete_event.Set();
-                                    }
-                                }};
+                if (current_status == Status::Initialized) {
+                    // Receiving data means the communication is ready now
+                    current_status = Status::Ready;
+                    status_callback(current_status);
+                }
+                const auto& touchpad_0 = data.touch[0];
+                if (touchpad_0.is_active == 0) {
+                    return;
+                }
+                LOG_DEBUG(Input, "Current touch: {} {}", touchpad_0.x, touchpad_0.y);
+                const u16 min_x = std::min(MAX_VALUE, static_cast<u16>(touchpad_0.x));
+                const u16 min_y = std::min(MAX_VALUE, static_cast<u16>(touchpad_0.y));
+                if (current_status == Status::Ready) {
+                    // First touch - min data (min_x/min_y)
+                    current_status = Status::Stage1Completed;
+                    status_callback(current_status);
+                }
+                if (touchpad_0.x - min_x > CALIBRATION_THRESHOLD &&
+                    touchpad_0.y - min_y > CALIBRATION_THRESHOLD) {
+                    // Set the current position as max value and finishes configuration
+                    const u16 max_x = touchpad_0.x;
+                    const u16 max_y = touchpad_0.y;
+                    current_status = Status::Completed;
+                    data_callback(min_x, min_y, max_x, max_y);
+                    status_callback(current_status);
+
+                    complete_event.Set();
+                }
+            }};
         Socket socket{host, port, std::move(callback)};
         std::thread worker_thread{SocketLoop, &socket};
         complete_event.Wait();

src/shader_recompiler/ir_opt/texture_pass.cpp

@@ -492,7 +492,8 @@ void TexturePass(Environment& env, IR::Program& program) {
             const auto insert_point{IR::Block::InstructionList::s_iterator_to(*inst)};
             IR::IREmitter ir{*texture_inst.block, insert_point};
             const IR::U32 shift{ir.Imm32(std::countr_zero(DESCRIPTOR_SIZE))};
-            inst->SetArg(0, ir.ShiftRightArithmetic(cbuf.dynamic_offset, shift));
+            inst->SetArg(0, ir.UMin(ir.ShiftRightArithmetic(cbuf.dynamic_offset, shift),
+                                    ir.Imm32(DESCRIPTOR_SIZE - 1)));
         } else {
             inst->SetArg(0, IR::Value{});
         }

src/video_core/dirty_flags.h

@@ -38,6 +38,9 @@ enum : u8 {
 
     Shaders,
 
+    // Special entries
+    DepthBiasGlobal,
+
     LastCommonEntry,
 };
 

src/video_core/gpu.h

@@ -216,7 +216,7 @@ public:
 
     void RegisterSyncptInterrupt(u32 syncpoint_id, u32 value);
 
-    [[nodiscard]] bool CancelSyncptInterrupt(u32 syncpoint_id, u32 value);
+    bool CancelSyncptInterrupt(u32 syncpoint_id, u32 value);
 
     [[nodiscard]] u64 GetTicks() const;
 

src/video_core/query_cache.h

@@ -258,9 +258,9 @@ private:
 
     void AsyncFlushQuery(VAddr addr) {
         if (!uncommitted_flushes) {
-            uncommitted_flushes = std::make_shared<std::unordered_set<VAddr>>();
+            uncommitted_flushes = std::make_shared<std::vector<VAddr>>();
         }
-        uncommitted_flushes->insert(addr);
+        uncommitted_flushes->push_back(addr);
     }
 
     static constexpr std::uintptr_t PAGE_SIZE = 4096;
@@ -276,8 +276,8 @@ private:
 
     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;
+    std::shared_ptr<std::vector<VAddr>> uncommitted_flushes{};
+    std::list<std::shared_ptr<std::vector<VAddr>>> committed_flushes;
 };
 
 template <class QueryCache, class HostCounter>

src/video_core/rasterizer_accelerated.h

@@ -42,7 +42,7 @@ private:
     };
     static_assert(sizeof(CacheEntry) == 8, "CacheEntry should be 8 bytes!");
 
-    std::array<CacheEntry, 0x1000000> cached_pages;
+    std::array<CacheEntry, 0x2000000> cached_pages;
     Core::Memory::Memory& cpu_memory;
 };
 

src/video_core/renderer_opengl/gl_device.cpp

@@ -10,6 +10,7 @@
 #include <limits>
 #include <optional>
 #include <span>
+#include <stdexcept>
 #include <vector>
 
 #include <glad/glad.h>

src/video_core/renderer_vulkan/vk_master_semaphore.h

@@ -21,12 +21,12 @@ public:
 
     /// Returns the current logical tick.
     [[nodiscard]] u64 CurrentTick() const noexcept {
-        return current_tick.load(std::memory_order_relaxed);
+        return current_tick.load(std::memory_order_acquire);
     }
 
     /// Returns the last known GPU tick.
     [[nodiscard]] u64 KnownGpuTick() const noexcept {
-        return gpu_tick.load(std::memory_order_relaxed);
+        return gpu_tick.load(std::memory_order_acquire);
     }
 
     /// Returns the timeline semaphore handle.
@@ -41,12 +41,21 @@ public:
 
     /// Advance to the logical tick and return the old one
     [[nodiscard]] u64 NextTick() noexcept {
-        return current_tick.fetch_add(1, std::memory_order::relaxed);
+        return current_tick.fetch_add(1, std::memory_order_release);
     }
 
     /// Refresh the known GPU tick
     void Refresh() {
-        gpu_tick.store(semaphore.GetCounter(), std::memory_order_relaxed);
+        u64 this_tick{};
+        u64 counter{};
+        do {
+            this_tick = gpu_tick.load(std::memory_order_acquire);
+            counter = semaphore.GetCounter();
+            if (counter < this_tick) {
+                return;
+            }
+        } while (!gpu_tick.compare_exchange_weak(this_tick, counter, std::memory_order_release,
+                                                 std::memory_order_relaxed));
     }
 
     /// Waits for a tick to be hit on the GPU

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -117,7 +117,8 @@ u64 HostCounter::BlockingQuery() const {
     cache.GetScheduler().Wait(tick);
     u64 data;
     const VkResult query_result = cache.GetDevice().GetLogical().GetQueryResults(
-        query.first, query.second, 1, sizeof(data), &data, sizeof(data), VK_QUERY_RESULT_64_BIT);
+        query.first, query.second, 1, sizeof(data), &data, sizeof(data),
+        VK_QUERY_RESULT_64_BIT | VK_QUERY_RESULT_WAIT_BIT);
 
     switch (query_result) {
     case VK_SUCCESS:

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -627,9 +627,21 @@ void RasterizerVulkan::UpdateDepthBias(Tegra::Engines::Maxwell3D::Regs& regs) {
     if (!state_tracker.TouchDepthBias()) {
         return;
     }
-    scheduler.Record([constant = regs.polygon_offset_units, clamp = regs.polygon_offset_clamp,
+    float units = regs.polygon_offset_units / 2.0f;
+    const bool is_d24 = regs.zeta.format == Tegra::DepthFormat::S8_UINT_Z24_UNORM ||
+                        regs.zeta.format == Tegra::DepthFormat::D24X8_UNORM ||
+                        regs.zeta.format == Tegra::DepthFormat::D24S8_UNORM ||
+                        regs.zeta.format == Tegra::DepthFormat::D24C8_UNORM;
+    if (is_d24 && !device.SupportsD24DepthBuffer()) {
+        // the base formulas can be obtained from here:
+        //   https://docs.microsoft.com/en-us/windows/win32/direct3d11/d3d10-graphics-programming-guide-output-merger-stage-depth-bias
+        const double rescale_factor =
+            static_cast<double>(1ULL << (32 - 24)) / (static_cast<double>(0x1.ep+127));
+        units = static_cast<float>(static_cast<double>(units) * rescale_factor);
+    }
+    scheduler.Record([constant = units, clamp = regs.polygon_offset_clamp,
                       factor = regs.polygon_offset_factor](vk::CommandBuffer cmdbuf) {
-        cmdbuf.SetDepthBias(constant, clamp, factor / 2.0f);
+        cmdbuf.SetDepthBias(constant, clamp, factor);
     });
 }
 

src/video_core/renderer_vulkan/vk_state_tracker.cpp

@@ -54,6 +54,7 @@ void SetupDirtyViewports(Tables& tables) {
     FillBlock(tables[0], OFF(viewport_transform), NUM(viewport_transform), Viewports);
     FillBlock(tables[0], OFF(viewports), NUM(viewports), Viewports);
     tables[0][OFF(viewport_transform_enabled)] = Viewports;
+    tables[1][OFF(screen_y_control)] = Viewports;
 }
 
 void SetupDirtyScissors(Tables& tables) {

src/video_core/renderer_vulkan/vk_state_tracker.h

@@ -79,7 +79,8 @@ public:
     }
 
     bool TouchDepthBias() {
-        return Exchange(Dirty::DepthBias, false);
+        return Exchange(Dirty::DepthBias, false) ||
+               Exchange(VideoCommon::Dirty::DepthBiasGlobal, false);
     }
 
     bool TouchBlendConstants() {

src/video_core/texture_cache/texture_cache.h

@@ -221,6 +221,7 @@ void TextureCache<P>::UpdateRenderTargets(bool is_clear) {
         BindRenderTarget(&render_targets.depth_buffer_id, FindDepthBuffer(is_clear));
     }
     const ImageViewId depth_buffer_id = render_targets.depth_buffer_id;
+
     PrepareImageView(depth_buffer_id, true, is_clear && IsFullClear(depth_buffer_id));
 
     for (size_t index = 0; index < NUM_RT; ++index) {
@@ -230,6 +231,8 @@ void TextureCache<P>::UpdateRenderTargets(bool is_clear) {
         maxwell3d.regs.render_area.width,
         maxwell3d.regs.render_area.height,
     };
+
+    flags[Dirty::DepthBiasGlobal] = true;
 }
 
 template <class P>

src/video_core/vulkan_common/vulkan_device.cpp

@@ -623,6 +623,10 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
         is_float16_supported = false;
     }
 
+    supports_d24_depth =
+        IsFormatSupported(VK_FORMAT_D24_UNORM_S8_UINT,
+                          VK_FORMAT_FEATURE_DEPTH_STENCIL_ATTACHMENT_BIT, FormatType::Optimal);
+
     graphics_queue = logical.GetQueue(graphics_family);
     present_queue = logical.GetQueue(present_family);
 }

src/video_core/vulkan_common/vulkan_device.h

@@ -332,6 +332,10 @@ public:
         return sets_per_pool;
     }
 
+    bool SupportsD24DepthBuffer() const {
+        return supports_d24_depth;
+    }
+
 private:
     /// Checks if the physical device is suitable.
     void CheckSuitability(bool requires_swapchain) const;
@@ -425,6 +429,7 @@ private:
     bool has_broken_cube_compatibility{};   ///< Has broken cube compatiblity bit
     bool has_renderdoc{};                   ///< Has RenderDoc attached
     bool has_nsight_graphics{};             ///< Has Nsight Graphics attached
+    bool supports_d24_depth{};              ///< Supports D24 depth buffers.
 
     // Telemetry parameters
     std::string vendor_name;                       ///< Device's driver name.

src/yuzu/configuration/config.cpp

@@ -918,8 +918,7 @@ void Config::ReadSystemValues() {
         const auto custom_rtc_enabled =
             ReadSetting(QStringLiteral("custom_rtc_enabled"), false).toBool();
         if (custom_rtc_enabled) {
-            Settings::values.custom_rtc =
-                std::chrono::seconds(ReadSetting(QStringLiteral("custom_rtc"), 0).toULongLong());
+            Settings::values.custom_rtc = ReadSetting(QStringLiteral("custom_rtc"), 0).toLongLong();
         } else {
             Settings::values.custom_rtc = std::nullopt;
         }
@@ -1450,9 +1449,7 @@ void Config::SaveSystemValues() {
         WriteSetting(QStringLiteral("custom_rtc_enabled"), Settings::values.custom_rtc.has_value(),
                      false);
         WriteSetting(QStringLiteral("custom_rtc"),
-                     QVariant::fromValue<long long>(
-                         Settings::values.custom_rtc.value_or(std::chrono::seconds{}).count()),
-                     0);
+                     QVariant::fromValue<long long>(Settings::values.custom_rtc.value_or(0)), 0);
     }
 
     WriteGlobalSetting(Settings::values.sound_index);

src/yuzu/configuration/configure_system.cpp

@@ -65,8 +65,7 @@ void ConfigureSystem::SetConfiguration() {
         QStringLiteral("%1")
             .arg(Settings::values.rng_seed.GetValue().value_or(0), 8, 16, QLatin1Char{'0'})
             .toUpper();
-    const auto rtc_time = Settings::values.custom_rtc.value_or(
-        std::chrono::seconds(QDateTime::currentSecsSinceEpoch()));
+    const auto rtc_time = Settings::values.custom_rtc.value_or(QDateTime::currentSecsSinceEpoch());
 
     ui->rng_seed_checkbox->setChecked(Settings::values.rng_seed.GetValue().has_value());
     ui->rng_seed_edit->setEnabled(Settings::values.rng_seed.GetValue().has_value() &&
@@ -75,7 +74,7 @@ void ConfigureSystem::SetConfiguration() {
 
     ui->custom_rtc_checkbox->setChecked(Settings::values.custom_rtc.has_value());
     ui->custom_rtc_edit->setEnabled(Settings::values.custom_rtc.has_value());
-    ui->custom_rtc_edit->setDateTime(QDateTime::fromSecsSinceEpoch(rtc_time.count()));
+    ui->custom_rtc_edit->setDateTime(QDateTime::fromSecsSinceEpoch(rtc_time));
 
     if (Settings::IsConfiguringGlobal()) {
         ui->combo_language->setCurrentIndex(Settings::values.language_index.GetValue());
@@ -108,10 +107,9 @@ void ConfigureSystem::ApplyConfiguration() {
     // to allow in-game time to be fast forwarded
     if (Settings::IsConfiguringGlobal()) {
         if (ui->custom_rtc_checkbox->isChecked()) {
-            Settings::values.custom_rtc =
-                std::chrono::seconds(ui->custom_rtc_edit->dateTime().toSecsSinceEpoch());
+            Settings::values.custom_rtc = ui->custom_rtc_edit->dateTime().toSecsSinceEpoch();
             if (system.IsPoweredOn()) {
-                const s64 posix_time{Settings::values.custom_rtc->count() +
+                const s64 posix_time{*Settings::values.custom_rtc +
                                      Service::Time::TimeManager::GetExternalTimeZoneOffset()};
                 system.GetTimeManager().UpdateLocalSystemClockTime(posix_time);
             }

src/yuzu/configuration/configure_tas.ui

@@ -14,14 +14,14 @@
         <item row="0" column="0" colspan="4">
          <widget class="QLabel" name="label_1">
           <property name="text">
-           <string>Reads controller input from scripts in the same format as TAS-nx scripts.&lt;br/&gt;For a more detailed explanation please consult the FAQ on the yuzu website.</string>
+           <string>&lt;html&gt;&lt;head/&gt;&lt;body&gt;&lt;p&gt;Reads controller input from scripts in the same format as TAS-nx scripts.&lt;br/&gt;For a more detailed explanation, please consult the &lt;a href=&quot;https://yuzu-emu.org/help/feature/tas/&quot;&gt;&lt;span style=&quot; text-decoration: underline; color:#039be5;&quot;&gt;help page&lt;/span&gt;&lt;/a&gt; on the yuzu website.&lt;/p&gt;&lt;/body&gt;&lt;/html&gt;</string>
           </property>
          </widget>
         </item>
         <item row="1" column="0" colspan="4">
          <widget class="QLabel" name="label_2">
           <property name="text">
-           <string>To check which hotkeys control the playback/recording, please refer to the Hotkey settings (General -&gt; Hotkeys).</string>
+           <string>To check which hotkeys control the playback/recording, please refer to the Hotkey settings (Configure -&gt; General -&gt; Hotkeys).</string>
           </property>
           <property name="wordWrap">
            <bool>true</bool>

src/yuzu/main.cpp

@@ -192,7 +192,7 @@ static void RemoveCachedContents() {
 }
 
 GMainWindow::GMainWindow()
-    : system{std::make_unique<Core::System>()},
+    : ui{std::make_unique<Ui::MainWindow>()}, system{std::make_unique<Core::System>()},
       input_subsystem{std::make_shared<InputCommon::InputSubsystem>()},
       config{std::make_unique<Config>(*system)},
       vfs{std::make_shared<FileSys::RealVfsFilesystem>()},
@@ -1666,7 +1666,7 @@ void GMainWindow::OnTransferableShaderCacheOpenFile(u64 program_id) {
     const auto shader_cache_folder_path{shader_cache_dir / fmt::format("{:016x}", program_id)};
     if (!Common::FS::CreateDirs(shader_cache_folder_path)) {
         QMessageBox::warning(this, tr("Error Opening Transferable Shader Cache"),
-                             tr("Filed to create the shader cache directory for this title."));
+                             tr("Failed to create the shader cache directory for this title."));
         return;
     }
     const auto shader_path_string{Common::FS::PathToUTF8String(shader_cache_folder_path)};

src/yuzu_cmd/config.cpp

@@ -412,8 +412,7 @@ void Config::ReadValues() {
 
     const auto custom_rtc_enabled = sdl2_config->GetBoolean("System", "custom_rtc_enabled", false);
     if (custom_rtc_enabled) {
-        Settings::values.custom_rtc =
-            std::chrono::seconds(sdl2_config->GetInteger("System", "custom_rtc", 0));
+        Settings::values.custom_rtc = sdl2_config->GetInteger("System", "custom_rtc", 0);
     } else {
         Settings::values.custom_rtc = std::nullopt;
     }