nvnflinger: release queued handles immediately on disconnection

kernel: refactor dummy thread wakeups

CMakeLists.txt

@@ -541,9 +541,9 @@ add_definitions(-DBOOST_ERROR_CODE_HEADER_ONLY
 # Adjustments for MSVC + Ninja
 if (MSVC AND CMAKE_GENERATOR STREQUAL "Ninja")
     add_compile_options(
-        /wd4711 # function 'function' selected for automatic inline expansion
         /wd4464 # relative include path contains '..'
-        /wd4820 # 'identifier1': '4' bytes padding added after data member 'identifier2'
+        /wd4711 # function 'function' selected for automatic inline expansion
+        /wd4820 # 'bytes' bytes padding added after construct 'member_name'
     )
 endif()
 

src/CMakeLists.txt

@@ -58,13 +58,11 @@ if (MSVC)
 
         # Warnings
         /W3
-        /we4018 # 'expression': signed/unsigned mismatch
+        /WX
+
         /we4062 # Enumerator 'identifier' in a switch of enum 'enumeration' is not handled
-        /we4101 # 'identifier': unreferenced local variable
         /we4189 # 'identifier': local variable is initialized but not referenced
         /we4265 # 'class': class has virtual functions, but destructor is not virtual
-        /we4267 # 'var': conversion from 'size_t' to 'type', possible loss of data
-        /we4305 # 'context': truncation from 'type1' to 'type2'
         /we4388 # 'expression': signed/unsigned mismatch
         /we4389 # 'operator': signed/unsigned mismatch
         /we4456 # Declaration of 'identifier' hides previous local declaration
@@ -75,10 +73,13 @@ if (MSVC)
         /we4547 # 'operator': operator before comma has no effect; expected operator with side-effect
         /we4549 # 'operator1': operator before comma has no effect; did you intend 'operator2'?
         /we4555 # Expression has no effect; expected expression with side-effect
-        /we4715 # 'function': not all control paths return a value
-        /we4834 # Discarding return value of function with 'nodiscard' attribute
+        /we4826 # Conversion from 'type1' to 'type2' is sign-extended. This may cause unexpected runtime behavior.
         /we5038 # data member 'member1' will be initialized after data member 'member2'
+        /we5233 # explicit lambda capture 'identifier' is not used
         /we5245 # 'function': unreferenced function with internal linkage has been removed
+
+        /wd4100 # 'identifier': unreferenced formal parameter
+        /wd4324 # 'struct_name': structure was padded due to __declspec(align())
     )
 
     if (USE_CCACHE)
@@ -99,24 +100,18 @@ if (MSVC)
     set(CMAKE_EXE_LINKER_FLAGS_RELEASE "/DEBUG /MANIFEST:NO /INCREMENTAL:NO /OPT:REF,ICF" CACHE STRING "" FORCE)
 else()
     add_compile_options(
-        -Wall
-        -Werror=array-bounds
-        -Werror=implicit-fallthrough
+        -Werror=all
+        -Werror=extra
         -Werror=missing-declarations
-        -Werror=missing-field-initializers
-        -Werror=reorder
         -Werror=shadow
-        -Werror=sign-compare
-        -Werror=switch
-        -Werror=uninitialized
-        -Werror=unused-function
-        -Werror=unused-result
-        -Werror=unused-variable
-        -Wextra
-        -Wmissing-declarations
+        -Werror=unused
+
         -Wno-attributes
         -Wno-invalid-offsetof
         -Wno-unused-parameter
+
+        $<$<CXX_COMPILER_ID:Clang>:-Wno-braced-scalar-init>
+        $<$<CXX_COMPILER_ID:Clang>:-Wno-unused-private-field>
     )
 
     if (ARCHITECTURE_x86_64)

src/audio_core/CMakeLists.txt

@@ -206,20 +206,11 @@ if (MSVC)
         /we4244 # 'conversion': conversion from 'type1' to 'type2', possible loss of data
         /we4245 # 'conversion': conversion from 'type1' to 'type2', signed/unsigned mismatch
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
-        /we4456 # Declaration of 'identifier' hides previous local declaration
-        /we4457 # Declaration of 'identifier' hides function parameter
-        /we4458 # Declaration of 'identifier' hides class member
-        /we4459 # Declaration of 'identifier' hides global declaration
+        /we4800 # Implicit conversion from 'type' to bool. Possible information loss
     )
 else()
     target_compile_options(audio_core PRIVATE
         -Werror=conversion
-        -Werror=ignored-qualifiers
-        -Werror=shadow
-        -Werror=unused-variable
-
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
 
         -Wno-sign-conversion
     )

src/audio_core/renderer/behavior/info_updater.cpp

@@ -91,7 +91,7 @@ Result InfoUpdater::UpdateVoices(VoiceContext& voice_context,
             voice_info.Initialize();
 
             for (u32 channel = 0; channel < in_param.channel_count; channel++) {
-                std::memset(voice_states[channel], 0, sizeof(VoiceState));
+                *voice_states[channel] = {};
             }
         }
 

src/audio_core/renderer/command/effect/biquad_filter.cpp

@@ -94,7 +94,7 @@ void BiquadFilterCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor
 void BiquadFilterCommand::Process(const ADSP::CommandListProcessor& processor) {
     auto state_{reinterpret_cast<VoiceState::BiquadFilterState*>(state)};
     if (needs_init) {
-        std::memset(state_, 0, sizeof(VoiceState::BiquadFilterState));
+        *state_ = {};
     }
 
     auto input_buffer{

src/audio_core/renderer/command/effect/multi_tap_biquad_filter.cpp

@@ -30,7 +30,7 @@ void MultiTapBiquadFilterCommand::Process(const ADSP::CommandListProcessor& proc
     for (u32 i = 0; i < filter_tap_count; i++) {
         auto state{reinterpret_cast<VoiceState::BiquadFilterState*>(states[i])};
         if (needs_init[i]) {
-            std::memset(state, 0, sizeof(VoiceState::BiquadFilterState));
+            *state = {};
         }
 
         ApplyBiquadFilterFloat(output_buffer, input_buffer, biquads[i].b, biquads[i].a, *state,

src/common/CMakeLists.txt

@@ -156,12 +156,13 @@ if (MSVC)
   )
   target_compile_options(common PRIVATE
     /W4
-    /WX
+
+    /we4242 # 'identifier': conversion from 'type1' to 'type2', possible loss of data
+    /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
+    /we4800 # Implicit conversion from 'type' to bool. Possible information loss
   )
 else()
   target_compile_options(common PRIVATE
-    -Werror
-
     $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>
   )
 endif()

src/common/bit_field.h

@@ -141,10 +141,6 @@ public:
     constexpr BitField(BitField&&) noexcept = default;
     constexpr BitField& operator=(BitField&&) noexcept = default;
 
-    [[nodiscard]] constexpr operator T() const {
-        return Value();
-    }
-
     constexpr void Assign(const T& value) {
 #ifdef _MSC_VER
         storage = static_cast<StorageType>((storage & ~mask) | FormatValue(value));
@@ -162,6 +158,17 @@ public:
         return ExtractValue(storage);
     }
 
+    template <typename ConvertedToType>
+    [[nodiscard]] constexpr ConvertedToType As() const {
+        static_assert(!std::is_same_v<T, ConvertedToType>,
+                      "Unnecessary cast. Use Value() instead.");
+        return static_cast<ConvertedToType>(Value());
+    }
+
+    [[nodiscard]] constexpr operator T() const {
+        return Value();
+    }
+
     [[nodiscard]] constexpr explicit operator bool() const {
         return Value() != 0;
     }

src/common/bounded_threadsafe_queue.h

@@ -21,11 +21,6 @@ constexpr size_t hardware_interference_size = std::hardware_destructive_interfer
 constexpr size_t hardware_interference_size = 64;
 #endif
 
-#ifdef _MSC_VER
-#pragma warning(push)
-#pragma warning(disable : 4324)
-#endif
-
 template <typename T, size_t capacity = 0x400>
 class MPSCQueue {
 public:
@@ -160,8 +155,4 @@ private:
     static_assert(std::is_nothrow_destructible_v<T>, "T must be nothrow destructible");
 };
 
-#ifdef _MSC_VER
-#pragma warning(pop)
-#endif
-
 } // namespace Common

src/common/concepts.h

@@ -3,24 +3,14 @@
 
 #pragma once
 
+#include <iterator>
 #include <type_traits>
 
 namespace Common {
 
-// Check if type is like an STL container
+// Check if type satisfies the ContiguousContainer named requirement.
 template <typename T>
-concept IsSTLContainer = requires(T t) {
-    typename T::value_type;
-    typename T::iterator;
-    typename T::const_iterator;
-    // TODO(ogniK): Replace below is std::same_as<void> when MSVC supports it.
-    t.begin();
-    t.end();
-    t.cbegin();
-    t.cend();
-    t.data();
-    t.size();
-};
+concept IsContiguousContainer = std::contiguous_iterator<typename T::iterator>;
 
 // TODO: Replace with std::derived_from when the <concepts> header
 //       is available on all supported platforms.

src/common/fs/file.h

@@ -209,8 +209,8 @@ public:
 
     /**
      * Helper function which deduces the value type of a contiguous STL container used in ReadSpan.
-     * If T is not a contiguous STL container as defined by the concept IsSTLContainer, this calls
-     * ReadObject and T must be a trivially copyable object.
+     * If T is not a contiguous container as defined by the concept IsContiguousContainer, this
+     * calls ReadObject and T must be a trivially copyable object.
      *
      * See ReadSpan for more details if T is a contiguous container.
      * See ReadObject for more details if T is a trivially copyable object.
@@ -223,7 +223,7 @@ public:
      */
     template <typename T>
     [[nodiscard]] size_t Read(T& data) const {
-        if constexpr (IsSTLContainer<T>) {
+        if constexpr (IsContiguousContainer<T>) {
             using ContiguousType = typename T::value_type;
             static_assert(std::is_trivially_copyable_v<ContiguousType>,
                           "Data type must be trivially copyable.");
@@ -235,8 +235,8 @@ public:
 
     /**
      * Helper function which deduces the value type of a contiguous STL container used in WriteSpan.
-     * If T is not a contiguous STL container as defined by the concept IsSTLContainer, this calls
-     * WriteObject and T must be a trivially copyable object.
+     * If T is not a contiguous STL container as defined by the concept IsContiguousContainer, this
+     * calls WriteObject and T must be a trivially copyable object.
      *
      * See WriteSpan for more details if T is a contiguous container.
      * See WriteObject for more details if T is a trivially copyable object.
@@ -249,7 +249,7 @@ public:
      */
     template <typename T>
     [[nodiscard]] size_t Write(const T& data) const {
-        if constexpr (IsSTLContainer<T>) {
+        if constexpr (IsContiguousContainer<T>) {
             using ContiguousType = typename T::value_type;
             static_assert(std::is_trivially_copyable_v<ContiguousType>,
                           "Data type must be trivially copyable.");

src/common/input.h

@@ -100,7 +100,6 @@ enum class CameraError {
 enum class VibrationAmplificationType {
     Linear,
     Exponential,
-    Test,
 };
 
 // Analog properties for calibration
@@ -325,6 +324,10 @@ public:
         return VibrationError::NotSupported;
     }
 
+    virtual bool IsVibrationEnabled() {
+        return false;
+    }
+
     virtual PollingError SetPollingMode([[maybe_unused]] PollingMode polling_mode) {
         return PollingError::NotSupported;
     }

src/core/CMakeLists.txt

@@ -774,19 +774,15 @@ if (MSVC)
         /we4244 # 'conversion': conversion from 'type1' to 'type2', possible loss of data
         /we4245 # 'conversion': conversion from 'type1' to 'type2', signed/unsigned mismatch
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
+        /we4800 # Implicit conversion from 'type' to bool. Possible information loss
     )
 else()
     target_compile_options(core PRIVATE
         -Werror=conversion
-        -Werror=ignored-qualifiers
-
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=class-memaccess>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
-
-        $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>
 
         -Wno-sign-conversion
+
+        $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>
     )
 endif()
 

src/core/core.cpp

@@ -384,6 +384,7 @@ struct System::Impl {
         kernel.ShutdownCores();
         cpu_manager.Shutdown();
         debugger.reset();
+        services->KillNVNFlinger();
         kernel.CloseServices();
         services.reset();
         service_manager.reset();

src/core/file_sys/card_image.cpp

@@ -232,8 +232,8 @@ const std::vector<std::shared_ptr<NCA>>& XCI::GetNCAs() const {
 
 std::shared_ptr<NCA> XCI::GetNCAByType(NCAContentType type) const {
     const auto program_id = secure_partition->GetProgramTitleID();
-    const auto iter = std::find_if(
-        ncas.begin(), ncas.end(), [this, type, program_id](const std::shared_ptr<NCA>& nca) {
+    const auto iter =
+        std::find_if(ncas.begin(), ncas.end(), [type, program_id](const std::shared_ptr<NCA>& nca) {
             return nca->GetType() == type && nca->GetTitleId() == program_id;
         });
     return iter == ncas.end() ? nullptr : *iter;

src/core/file_sys/program_metadata.cpp

@@ -127,7 +127,7 @@ void ProgramMetadata::LoadManual(bool is_64_bit, ProgramAddressSpaceType address
 }
 
 bool ProgramMetadata::Is64BitProgram() const {
-    return npdm_header.has_64_bit_instructions;
+    return npdm_header.has_64_bit_instructions.As<bool>();
 }
 
 ProgramAddressSpaceType ProgramMetadata::GetAddressSpaceType() const {

src/core/hid/emulated_controller.cpp

@@ -970,14 +970,7 @@ bool EmulatedController::SetVibration(std::size_t device_index, VibrationValue v
            Common::Input::VibrationError::None;
 }
 
-bool EmulatedController::TestVibration(std::size_t device_index) {
-    if (device_index >= output_devices.size()) {
-        return false;
-    }
-    if (!output_devices[device_index]) {
-        return false;
-    }
-
+bool EmulatedController::IsVibrationEnabled(std::size_t device_index) {
     const auto player_index = NpadIdTypeToIndex(npad_id_type);
     const auto& player = Settings::values.players.GetValue()[player_index];
 
@@ -985,31 +978,15 @@ bool EmulatedController::TestVibration(std::size_t device_index) {
         return false;
     }
 
-    const Common::Input::VibrationStatus test_vibration = {
-        .low_amplitude = 0.001f,
-        .low_frequency = DEFAULT_VIBRATION_VALUE.low_frequency,
-        .high_amplitude = 0.001f,
-        .high_frequency = DEFAULT_VIBRATION_VALUE.high_frequency,
-        .type = Common::Input::VibrationAmplificationType::Test,
-    };
+    if (device_index >= output_devices.size()) {
+        return false;
+    }
 
-    const Common::Input::VibrationStatus zero_vibration = {
-        .low_amplitude = DEFAULT_VIBRATION_VALUE.low_amplitude,
-        .low_frequency = DEFAULT_VIBRATION_VALUE.low_frequency,
-        .high_amplitude = DEFAULT_VIBRATION_VALUE.high_amplitude,
-        .high_frequency = DEFAULT_VIBRATION_VALUE.high_frequency,
-        .type = Common::Input::VibrationAmplificationType::Test,
-    };
+    if (!output_devices[device_index]) {
+        return false;
+    }
 
-    // Send a slight vibration to test for rumble support
-    output_devices[device_index]->SetVibration(test_vibration);
-
-    // Wait for about 15ms to ensure the controller is ready for the stop command
-    std::this_thread::sleep_for(std::chrono::milliseconds(15));
-
-    // Stop any vibration and return the result
-    return output_devices[device_index]->SetVibration(zero_vibration) ==
-           Common::Input::VibrationError::None;
+    return output_devices[device_index]->IsVibrationEnabled();
 }
 
 bool EmulatedController::SetPollingMode(Common::Input::PollingMode polling_mode) {
@@ -1048,6 +1025,7 @@ bool EmulatedController::HasNfc() const {
     case NpadStyleIndex::JoyconRight:
     case NpadStyleIndex::JoyconDual:
     case NpadStyleIndex::ProController:
+    case NpadStyleIndex::Handheld:
         break;
     default:
         return false;
@@ -1158,27 +1136,27 @@ bool EmulatedController::IsControllerSupported(bool use_temporary_value) const {
     const auto type = is_configuring && use_temporary_value ? tmp_npad_type : npad_type;
     switch (type) {
     case NpadStyleIndex::ProController:
-        return supported_style_tag.fullkey;
+        return supported_style_tag.fullkey.As<bool>();
     case NpadStyleIndex::Handheld:
-        return supported_style_tag.handheld;
+        return supported_style_tag.handheld.As<bool>();
     case NpadStyleIndex::JoyconDual:
-        return supported_style_tag.joycon_dual;
+        return supported_style_tag.joycon_dual.As<bool>();
     case NpadStyleIndex::JoyconLeft:
-        return supported_style_tag.joycon_left;
+        return supported_style_tag.joycon_left.As<bool>();
     case NpadStyleIndex::JoyconRight:
-        return supported_style_tag.joycon_right;
+        return supported_style_tag.joycon_right.As<bool>();
     case NpadStyleIndex::GameCube:
-        return supported_style_tag.gamecube;
+        return supported_style_tag.gamecube.As<bool>();
     case NpadStyleIndex::Pokeball:
-        return supported_style_tag.palma;
+        return supported_style_tag.palma.As<bool>();
     case NpadStyleIndex::NES:
-        return supported_style_tag.lark;
+        return supported_style_tag.lark.As<bool>();
     case NpadStyleIndex::SNES:
-        return supported_style_tag.lucia;
+        return supported_style_tag.lucia.As<bool>();
     case NpadStyleIndex::N64:
-        return supported_style_tag.lagoon;
+        return supported_style_tag.lagoon.As<bool>();
     case NpadStyleIndex::SegaGenesis:
-        return supported_style_tag.lager;
+        return supported_style_tag.lager.As<bool>();
     default:
         return false;
     }
@@ -1234,12 +1212,6 @@ bool EmulatedController::IsConnected(bool get_temporary_value) const {
     return is_connected;
 }
 
-bool EmulatedController::IsVibrationEnabled() const {
-    const auto player_index = NpadIdTypeToIndex(npad_id_type);
-    const auto& player = Settings::values.players.GetValue()[player_index];
-    return player.vibration_enabled;
-}
-
 NpadIdType EmulatedController::GetNpadIdType() const {
     std::scoped_lock lock{mutex};
     return npad_id_type;

src/core/hid/emulated_controller.h

@@ -206,9 +206,6 @@ public:
      */
     bool IsConnected(bool get_temporary_value = false) const;
 
-    /// Returns true if vibration is enabled
-    bool IsVibrationEnabled() const;
-
     /// Removes all callbacks created from input devices
     void UnloadInput();
 
@@ -339,7 +336,7 @@ public:
      * Sends a small vibration to the output device
      * @return true if SetVibration was successfull
      */
-    bool TestVibration(std::size_t device_index);
+    bool IsVibrationEnabled(std::size_t device_index);
 
     /**
      * Sets the desired data to be polled from a controller

src/core/hle/ipc_helpers.h

@@ -406,7 +406,7 @@ inline s32 RequestParser::Pop() {
 }
 
 // Ignore the -Wclass-memaccess warning on memcpy for non-trivially default constructible objects.
-#if defined(__GNUC__)
+#if defined(__GNUC__) && !defined(__clang__) && !defined(__INTEL_COMPILER)
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wclass-memaccess"
 #endif
@@ -417,7 +417,7 @@ void RequestParser::PopRaw(T& value) {
     std::memcpy(&value, cmdbuf + index, sizeof(T));
     index += (sizeof(T) + 3) / 4; // round up to word length
 }
-#if defined(__GNUC__)
+#if defined(__GNUC__) && !defined(__clang__) && !defined(__INTEL_COMPILER)
 #pragma GCC diagnostic pop
 #endif
 

src/core/hle/kernel/global_scheduler_context.cpp

@@ -49,4 +49,26 @@ bool GlobalSchedulerContext::IsLocked() const {
     return scheduler_lock.IsLockedByCurrentThread();
 }
 
+void GlobalSchedulerContext::RegisterDummyThreadForWakeup(KThread* thread) {
+    ASSERT(IsLocked());
+
+    woken_dummy_threads.insert(thread);
+}
+
+void GlobalSchedulerContext::UnregisterDummyThreadForWakeup(KThread* thread) {
+    ASSERT(IsLocked());
+
+    woken_dummy_threads.erase(thread);
+}
+
+void GlobalSchedulerContext::WakeupWaitingDummyThreads() {
+    ASSERT(IsLocked());
+
+    for (auto* thread : woken_dummy_threads) {
+        thread->DummyThreadEndWait();
+    }
+
+    woken_dummy_threads.clear();
+}
+
 } // namespace Kernel

src/core/hle/kernel/global_scheduler_context.h

@@ -4,6 +4,7 @@
 #pragma once
 
 #include <atomic>
+#include <set>
 #include <vector>
 
 #include "common/common_types.h"
@@ -58,6 +59,10 @@ public:
     /// Returns true if the global scheduler lock is acquired
     bool IsLocked() const;
 
+    void UnregisterDummyThreadForWakeup(KThread* thread);
+    void RegisterDummyThreadForWakeup(KThread* thread);
+    void WakeupWaitingDummyThreads();
+
     [[nodiscard]] LockType& SchedulerLock() {
         return scheduler_lock;
     }
@@ -76,6 +81,9 @@ private:
     KSchedulerPriorityQueue priority_queue;
     LockType scheduler_lock;
 
+    /// Lists dummy threads pending wakeup on lock release
+    std::set<KThread*> woken_dummy_threads;
+
     /// Lists all thread ids that aren't deleted/etc.
     std::vector<KThread*> thread_list;
     std::mutex global_list_guard;

src/core/hle/kernel/hle_ipc.h

@@ -304,7 +304,7 @@ public:
      */
     template <typename T, typename = std::enable_if_t<!std::is_pointer_v<T>>>
     std::size_t WriteBuffer(const T& data, std::size_t buffer_index = 0) const {
-        if constexpr (Common::IsSTLContainer<T>) {
+        if constexpr (Common::IsContiguousContainer<T>) {
             using ContiguousType = typename T::value_type;
             static_assert(std::is_trivially_copyable_v<ContiguousType>,
                           "Container to WriteBuffer must contain trivially copyable objects");

src/core/hle/kernel/k_scheduler.cpp

@@ -81,8 +81,8 @@ void KScheduler::RescheduleCurrentHLEThread(KernelCore& kernel) {
     // HACK: we cannot schedule from this thread, it is not a core thread
     ASSERT(GetCurrentThread(kernel).GetDisableDispatchCount() == 1);
 
-    // Special case to ensure dummy threads that are waiting block
-    GetCurrentThread(kernel).IfDummyThreadTryWait();
+    // Ensure dummy threads that are waiting block.
+    GetCurrentThread(kernel).DummyThreadBeginWait();
 
     ASSERT(GetCurrentThread(kernel).GetState() != ThreadState::Waiting);
     GetCurrentThread(kernel).EnableDispatch();
@@ -314,6 +314,16 @@ u64 KScheduler::UpdateHighestPriorityThreadsImpl(KernelCore& kernel) {
         idle_cores &= ~(1ULL << core_id);
     }
 
+    // HACK: any waiting dummy threads can wake up now.
+    kernel.GlobalSchedulerContext().WakeupWaitingDummyThreads();
+
+    // HACK: if we are a dummy thread, and we need to go sleep, indicate
+    // that for when the lock is released.
+    KThread* const cur_thread = GetCurrentThreadPointer(kernel);
+    if (cur_thread->IsDummyThread() && cur_thread->GetState() != ThreadState::Runnable) {
+        cur_thread->RequestDummyThreadWait();
+    }
+
     return cores_needing_scheduling;
 }
 
@@ -531,11 +541,23 @@ void KScheduler::OnThreadStateChanged(KernelCore& kernel, KThread* thread, Threa
         GetPriorityQueue(kernel).Remove(thread);
         IncrementScheduledCount(thread);
         SetSchedulerUpdateNeeded(kernel);
+
+        if (thread->IsDummyThread()) {
+            // HACK: if this is a dummy thread, it should no longer wake up when the
+            // scheduler lock is released.
+            kernel.GlobalSchedulerContext().UnregisterDummyThreadForWakeup(thread);
+        }
     } else if (cur_state == ThreadState::Runnable) {
         // If we're now runnable, then we weren't previously, and we should add.
         GetPriorityQueue(kernel).PushBack(thread);
         IncrementScheduledCount(thread);
         SetSchedulerUpdateNeeded(kernel);
+
+        if (thread->IsDummyThread()) {
+            // HACK: if this is a dummy thread, it should wake up when the scheduler
+            // lock is released.
+            kernel.GlobalSchedulerContext().RegisterDummyThreadForWakeup(thread);
+        }
     }
 }
 

src/core/hle/kernel/k_thread.cpp

@@ -148,7 +148,9 @@ Result KThread::Initialize(KThreadFunction func, uintptr_t arg, VAddr user_stack
     physical_affinity_mask.SetAffinity(phys_core, true);
 
     // Set the thread state.
-    thread_state = (type == ThreadType::Main) ? ThreadState::Runnable : ThreadState::Initialized;
+    thread_state = (type == ThreadType::Main || type == ThreadType::Dummy)
+                       ? ThreadState::Runnable
+                       : ThreadState::Initialized;
 
     // Set TLS address.
     tls_address = 0;
@@ -1174,30 +1176,29 @@ Result KThread::Sleep(s64 timeout) {
     R_SUCCEED();
 }
 
-void KThread::IfDummyThreadTryWait() {
-    if (!IsDummyThread()) {
-        return;
-    }
+void KThread::RequestDummyThreadWait() {
+    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(kernel));
+    ASSERT(this->IsDummyThread());
 
-    if (GetState() != ThreadState::Waiting) {
-        return;
-    }
-
-    ASSERT(!kernel.IsPhantomModeForSingleCore());
-
-    // Block until we are no longer waiting.
-    std::unique_lock lk(dummy_wait_lock);
-    dummy_wait_cv.wait(
-        lk, [&] { return GetState() != ThreadState::Waiting || kernel.IsShuttingDown(); });
+    // We will block when the scheduler lock is released.
+    dummy_thread_runnable.store(false);
 }
 
-void KThread::IfDummyThreadEndWait() {
-    if (!IsDummyThread()) {
-        return;
-    }
+void KThread::DummyThreadBeginWait() {
+    ASSERT(this->IsDummyThread());
+    ASSERT(!kernel.IsPhantomModeForSingleCore());
+
+    // Block until runnable is no longer false.
+    dummy_thread_runnable.wait(false);
+}
+
+void KThread::DummyThreadEndWait() {
+    ASSERT(KScheduler::IsSchedulerLockedByCurrentThread(kernel));
+    ASSERT(this->IsDummyThread());
 
     // Wake up the waiting thread.
-    dummy_wait_cv.notify_one();
+    dummy_thread_runnable.store(true);
+    dummy_thread_runnable.notify_one();
 }
 
 void KThread::BeginWait(KThreadQueue* queue) {
@@ -1231,9 +1232,6 @@ void KThread::EndWait(Result wait_result_) {
         }
 
         wait_queue->EndWait(this, wait_result_);
-
-        // Special case for dummy threads to wakeup if necessary.
-        IfDummyThreadEndWait();
     }
 }
 

src/core/hle/kernel/k_thread.h

@@ -643,8 +643,9 @@ public:
     // therefore will not block on guest kernel synchronization primitives. These methods handle
     // blocking as needed.
 
-    void IfDummyThreadTryWait();
-    void IfDummyThreadEndWait();
+    void RequestDummyThreadWait();
+    void DummyThreadBeginWait();
+    void DummyThreadEndWait();
 
     [[nodiscard]] uintptr_t GetArgument() const {
         return argument;
@@ -777,8 +778,7 @@ private:
     bool is_single_core{};
     ThreadType thread_type{};
     StepState step_state{};
-    std::mutex dummy_wait_lock;
-    std::condition_variable dummy_wait_cv;
+    std::atomic<bool> dummy_thread_runnable{true};
 
     // For debugging
     std::vector<KSynchronizationObject*> wait_objects_for_debugging;

src/core/hle/kernel/kernel.cpp

@@ -48,8 +48,8 @@ namespace Kernel {
 
 struct KernelCore::Impl {
     explicit Impl(Core::System& system_, KernelCore& kernel_)
-        : time_manager{system_},
-          service_threads_manager{1, "ServiceThreadsManager"}, system{system_} {}
+        : time_manager{system_}, service_threads_manager{1, "ServiceThreadsManager"},
+          service_thread_barrier{2}, system{system_} {}
 
     void SetMulticore(bool is_multi) {
         is_multicore = is_multi;
@@ -737,7 +737,12 @@ struct KernelCore::Impl {
     }
 
     void ClearServiceThreads() {
-        service_threads_manager.QueueWork([this]() { service_threads.clear(); });
+        service_threads_manager.QueueWork([this] {
+            service_threads.clear();
+            default_service_thread.reset();
+            service_thread_barrier.Sync();
+        });
+        service_thread_barrier.Sync();
     }
 
     std::mutex server_objects_lock;
@@ -802,6 +807,7 @@ struct KernelCore::Impl {
     std::unordered_set<std::shared_ptr<ServiceThread>> service_threads;
     std::weak_ptr<ServiceThread> default_service_thread;
     Common::ThreadWorker service_threads_manager;
+    Common::Barrier service_thread_barrier;
 
     std::array<KThread*, Core::Hardware::NUM_CPU_CORES> shutdown_threads;
     std::array<std::unique_ptr<Kernel::KScheduler>, Core::Hardware::NUM_CPU_CORES> schedulers{};

src/core/hle/kernel/svc.cpp

@@ -751,8 +751,8 @@ static void Break(Core::System& system, u32 reason, u64 info1, u64 info2) {
     }
 
     system.GetReporter().SaveSvcBreakReport(
-        static_cast<u32>(break_reason.break_type.Value()), break_reason.signal_debugger, info1,
-        info2, has_dumped_buffer ? std::make_optional(debug_buffer) : std::nullopt);
+        static_cast<u32>(break_reason.break_type.Value()), break_reason.signal_debugger.As<bool>(),
+        info1, info2, has_dumped_buffer ? std::make_optional(debug_buffer) : std::nullopt);
 
     if (!break_reason.signal_debugger) {
         LOG_CRITICAL(

src/core/hle/service/am/applets/applets.h

@@ -164,7 +164,7 @@ protected:
         u32_le size;
         u32_le library_version;
         u32_le theme_color;
-        u8 play_startup_sound;
+        bool play_startup_sound;
         u64_le system_tick;
     };
     static_assert(sizeof(CommonArguments) == 0x20, "CommonArguments has incorrect size.");

src/core/hle/service/hid/controllers/npad.cpp

@@ -868,7 +868,7 @@ bool Controller_NPad::VibrateControllerAtIndex(Core::HID::NpadIdType npad_id,
         return false;
     }
 
-    if (!controller.device->IsVibrationEnabled()) {
+    if (!controller.device->IsVibrationEnabled(device_index)) {
         if (controller.vibration[device_index].latest_vibration_value.low_amplitude != 0.0f ||
             controller.vibration[device_index].latest_vibration_value.high_amplitude != 0.0f) {
             // Send an empty vibration to stop any vibrations.
@@ -1001,7 +1001,7 @@ void Controller_NPad::InitializeVibrationDeviceAtIndex(Core::HID::NpadIdType npa
     }
 
     controller.vibration[device_index].device_mounted =
-        controller.device->TestVibration(device_index);
+        controller.device->IsVibrationEnabled(device_index);
 }
 
 void Controller_NPad::SetPermitVibrationSession(bool permit_vibration_session) {
@@ -1502,25 +1502,25 @@ bool Controller_NPad::IsControllerSupported(Core::HID::NpadStyleIndex controller
         Core::HID::NpadStyleTag style = GetSupportedStyleSet();
         switch (controller) {
         case Core::HID::NpadStyleIndex::ProController:
-            return style.fullkey;
+            return style.fullkey.As<bool>();
         case Core::HID::NpadStyleIndex::JoyconDual:
-            return style.joycon_dual;
+            return style.joycon_dual.As<bool>();
         case Core::HID::NpadStyleIndex::JoyconLeft:
-            return style.joycon_left;
+            return style.joycon_left.As<bool>();
         case Core::HID::NpadStyleIndex::JoyconRight:
-            return style.joycon_right;
+            return style.joycon_right.As<bool>();
         case Core::HID::NpadStyleIndex::GameCube:
-            return style.gamecube;
+            return style.gamecube.As<bool>();
         case Core::HID::NpadStyleIndex::Pokeball:
-            return style.palma;
+            return style.palma.As<bool>();
         case Core::HID::NpadStyleIndex::NES:
-            return style.lark;
+            return style.lark.As<bool>();
         case Core::HID::NpadStyleIndex::SNES:
-            return style.lucia;
+            return style.lucia.As<bool>();
         case Core::HID::NpadStyleIndex::N64:
-            return style.lagoon;
+            return style.lagoon.As<bool>();
         case Core::HID::NpadStyleIndex::SegaGenesis:
-            return style.lager;
+            return style.lager.As<bool>();
         default:
             return false;
         }

src/core/hle/service/nfp/amiibo_crypto.cpp

@@ -9,6 +9,7 @@
 #include <mbedtls/hmac_drbg.h>
 
 #include "common/fs/file.h"
+#include "common/fs/fs.h"
 #include "common/fs/path_util.h"
 #include "common/logging/log.h"
 #include "core/hle/service/mii/mii_manager.h"
@@ -279,7 +280,7 @@ bool LoadKeys(InternalKey& locked_secret, InternalKey& unfixed_info) {
                                        Common::FS::FileType::BinaryFile};
 
     if (!keys_file.IsOpen()) {
-        LOG_ERROR(Service_NFP, "No keys detected");
+        LOG_ERROR(Service_NFP, "Failed to open key file");
         return false;
     }
 
@@ -295,6 +296,11 @@ bool LoadKeys(InternalKey& locked_secret, InternalKey& unfixed_info) {
     return true;
 }
 
+bool IsKeyAvailable() {
+    const auto yuzu_keys_dir = Common::FS::GetYuzuPath(Common::FS::YuzuPath::KeysDir);
+    return Common::FS::Exists(yuzu_keys_dir / "key_retail.bin");
+}
+
 bool DecodeAmiibo(const EncryptedNTAG215File& encrypted_tag_data, NTAG215File& tag_data) {
     InternalKey locked_secret{};
     InternalKey unfixed_info{};

src/core/hle/service/nfp/amiibo_crypto.h

@@ -91,6 +91,9 @@ void Cipher(const DerivedKeys& keys, const NTAG215File& in_data, NTAG215File& ou
 /// Loads both amiibo keys from key_retail.bin
 bool LoadKeys(InternalKey& locked_secret, InternalKey& unfixed_info);
 
+/// Returns true if key_retail.bin exist
+bool IsKeyAvailable();
+
 /// Decodes encripted amiibo data returns true if output is valid
 bool DecodeAmiibo(const EncryptedNTAG215File& encrypted_tag_data, NTAG215File& tag_data);
 

src/core/hle/service/nfp/nfp_device.cpp

@@ -17,6 +17,7 @@
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/service/mii/mii_manager.h"
+#include "core/hle/service/mii/types.h"
 #include "core/hle/service/nfp/amiibo_crypto.h"
 #include "core/hle/service/nfp/nfp.h"
 #include "core/hle/service/nfp/nfp_device.h"
@@ -233,6 +234,14 @@ Result NfpDevice::Mount(MountTarget mount_target_) {
         return NotAnAmiibo;
     }
 
+    // Mark amiibos as read only when keys are missing
+    if (!AmiiboCrypto::IsKeyAvailable()) {
+        LOG_ERROR(Service_NFP, "No keys detected");
+        device_state = DeviceState::TagMounted;
+        mount_target = MountTarget::Rom;
+        return ResultSuccess;
+    }
+
     if (!AmiiboCrypto::DecodeAmiibo(encrypted_tag_data, tag_data)) {
         LOG_ERROR(Service_NFP, "Can't decode amiibo {}", device_state);
         return CorruptedData;

src/core/hle/service/nfp/nfp_device.h

@@ -8,7 +8,6 @@
 
 #include "common/common_funcs.h"
 #include "core/hle/service/kernel_helpers.h"
-#include "core/hle/service/mii/types.h"
 #include "core/hle/service/nfp/nfp_types.h"
 #include "core/hle/service/service.h"
 

src/core/hle/service/nfp/nfp_types.h

@@ -17,11 +17,6 @@ enum class ServiceType : u32 {
     System,
 };
 
-enum class State : u32 {
-    NonInitialized,
-    Initialized,
-};
-
 enum class DeviceState : u32 {
     Initialized,
     SearchingForTag,

src/core/hle/service/nfp/nfp_user.cpp

@@ -6,12 +6,9 @@
 
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/hid/emulated_controller.h"
-#include "core/hid/hid_core.h"
 #include "core/hid/hid_types.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/k_event.h"
-#include "core/hle/service/mii/mii_manager.h"
 #include "core/hle/service/nfp/nfp_device.h"
 #include "core/hle/service/nfp/nfp_result.h"
 #include "core/hle/service/nfp/nfp_user.h"

src/core/hle/service/nfp/nfp_user.h

@@ -4,8 +4,7 @@
 #pragma once
 
 #include "core/hle/service/kernel_helpers.h"
-#include "core/hle/service/nfp/nfp.h"
-#include "core/hle/service/nfp/nfp_types.h"
+#include "core/hle/service/service.h"
 
 namespace Service::NFP {
 class NfpDevice;
@@ -15,6 +14,11 @@ public:
     explicit IUser(Core::System& system_);
 
 private:
+    enum class State : u32 {
+        NonInitialized,
+        Initialized,
+    };
+
     void Initialize(Kernel::HLERequestContext& ctx);
     void Finalize(Kernel::HLERequestContext& ctx);
     void ListDevices(Kernel::HLERequestContext& ctx);

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

@@ -255,15 +255,16 @@ std::optional<NvMap::FreeInfo> NvMap::FreeHandle(Handle::Id handle, bool interna
             .address = handle_description->address,
             .size = handle_description->size,
             .was_uncached = handle_description->flags.map_uncached.Value() != 0,
+            .can_unlock = true,
         };
     } else {
         return std::nullopt;
     }
 
-    // Handle hasn't been freed from memory, set address to 0 to mark that the handle wasn't freed
+    // If the handle hasn't been freed from memory, mark that
     if (!hWeak.expired()) {
         LOG_DEBUG(Service_NVDRV, "nvmap handle: {} wasn't freed as it is still in use", handle);
-        freeInfo.address = 0;
+        freeInfo.can_unlock = false;
     }
 
     return freeInfo;

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

@@ -105,6 +105,7 @@ public:
         u64 address;       //!< Address the handle referred to before deletion
         u64 size;          //!< Page-aligned handle size
         bool was_uncached; //!< If the handle was allocated as uncached
+        bool can_unlock;   //!< If the address region is ready to be unlocked
     };
 
     explicit NvMap(Tegra::Host1x::Host1x& host1x);

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

@@ -251,10 +251,12 @@ NvResult nvmap::IocFree(const std::vector<u8>& input, std::vector<u8>& output) {
     }
 
     if (auto freeInfo{file.FreeHandle(params.handle, false)}) {
-        ASSERT(system.CurrentProcess()
-                   ->PageTable()
-                   .UnlockForDeviceAddressSpace(freeInfo->address, freeInfo->size)
-                   .IsSuccess());
+        if (freeInfo->can_unlock) {
+            ASSERT(system.CurrentProcess()
+                       ->PageTable()
+                       .UnlockForDeviceAddressSpace(freeInfo->address, freeInfo->size)
+                       .IsSuccess());
+        }
         params.address = freeInfo->address;
         params.size = static_cast<u32>(freeInfo->size);
         params.flags.raw = 0;

src/core/hle/service/nvflinger/buffer_queue_producer.cpp

@@ -742,6 +742,13 @@ Status BufferQueueProducer::Disconnect(NativeWindowApi api) {
             return Status::NoError;
         }
 
+        // HACK: We are not Android. Remove handle for items in queue, and clear queue.
+        // Allows synchronous destruction of nvmap handles.
+        for (auto& item : core->queue) {
+            nvmap.FreeHandle(item.graphic_buffer->BufferId(), true);
+        }
+        core->queue.clear();
+
         switch (api) {
         case NativeWindowApi::Egl:
         case NativeWindowApi::Cpu:

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -102,15 +102,19 @@ NVFlinger::~NVFlinger() {
         system.CoreTiming().UnscheduleEvent(single_composition_event, {});
     }
 
+    ShutdownLayers();
+
+    if (nvdrv) {
+        nvdrv->Close(disp_fd);
+    }
+}
+
+void NVFlinger::ShutdownLayers() {
     for (auto& display : displays) {
         for (size_t layer = 0; layer < display.GetNumLayers(); ++layer) {
             display.GetLayer(layer).Core().NotifyShutdown();
         }
     }
-
-    if (nvdrv) {
-        nvdrv->Close(disp_fd);
-    }
 }
 
 void NVFlinger::SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance) {
@@ -134,6 +138,19 @@ std::optional<u64> NVFlinger::OpenDisplay(std::string_view name) {
     return itr->GetID();
 }
 
+bool NVFlinger::CloseDisplay(u64 display_id) {
+    const auto lock_guard = Lock();
+    auto* const display = FindDisplay(display_id);
+
+    if (display == nullptr) {
+        return false;
+    }
+
+    display->Reset();
+
+    return true;
+}
+
 std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
     const auto lock_guard = Lock();
     auto* const display = FindDisplay(display_id);

src/core/hle/service/nvflinger/nvflinger.h

@@ -48,6 +48,8 @@ public:
     explicit NVFlinger(Core::System& system_, HosBinderDriverServer& hos_binder_driver_server_);
     ~NVFlinger();
 
+    void ShutdownLayers();
+
     /// Sets the NVDrv module instance to use to send buffers to the GPU.
     void SetNVDrvInstance(std::shared_ptr<Nvidia::Module> instance);
 
@@ -56,6 +58,11 @@ public:
     /// If an invalid display name is provided, then an empty optional is returned.
     [[nodiscard]] std::optional<u64> OpenDisplay(std::string_view name);
 
+    /// Closes the specified display by its ID.
+    ///
+    /// Returns false if an invalid display ID is provided.
+    [[nodiscard]] bool CloseDisplay(u64 display_id);
+
     /// Creates a layer on the specified display and returns the layer ID.
     ///
     /// If an invalid display ID is specified, then an empty optional is returned.

src/core/hle/service/service.cpp

@@ -303,4 +303,8 @@ Services::Services(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system
 
 Services::~Services() = default;
 
+void Services::KillNVNFlinger() {
+    nv_flinger->ShutdownLayers();
+}
+
 } // namespace Service

src/core/hle/service/service.h

@@ -238,6 +238,8 @@ public:
     explicit Services(std::shared_ptr<SM::ServiceManager>& sm, Core::System& system);
     ~Services();
 
+    void KillNVNFlinger();
+
 private:
     std::unique_ptr<NVFlinger::HosBinderDriverServer> hos_binder_driver_server;
     std::unique_ptr<NVFlinger::NVFlinger> nv_flinger;

src/core/hle/service/vi/display/vi_display.h

@@ -106,6 +106,12 @@ public:
     ///
     void CloseLayer(u64 layer_id);
 
+    /// Resets the display for a new connection.
+    void Reset() {
+        layers.clear();
+        got_vsync_event = false;
+    }
+
     /// Attempts to find a layer with the given ID.
     ///
     /// @param layer_id The layer ID.

src/core/hle/service/vi/vi.cpp

@@ -324,10 +324,10 @@ private:
         IPC::RequestParser rp{ctx};
         const u64 display = rp.Pop<u64>();
 
-        LOG_WARNING(Service_VI, "(STUBBED) called. display=0x{:016X}", display);
+        const Result rc = nv_flinger.CloseDisplay(display) ? ResultSuccess : ResultUnknown;
 
         IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
+        rb.Push(rc);
     }
 
     void CreateManagedLayer(Kernel::HLERequestContext& ctx) {
@@ -508,10 +508,10 @@ private:
         IPC::RequestParser rp{ctx};
         const u64 display_id = rp.Pop<u64>();
 
-        LOG_WARNING(Service_VI, "(STUBBED) called. display_id=0x{:016X}", display_id);
+        const Result rc = nv_flinger.CloseDisplay(display_id) ? ResultSuccess : ResultUnknown;
 
         IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
+        rb.Push(rc);
     }
 
     // This literally does nothing internally in the actual service itself,

src/core/memory.cpp

@@ -233,18 +233,17 @@ struct Memory::Impl {
                           current_vaddr, src_addr, size);
                 std::memset(dest_buffer, 0, copy_amount);
             },
-            [&dest_buffer](const std::size_t copy_amount, const u8* const src_ptr) {
+            [&](const std::size_t copy_amount, const u8* const src_ptr) {
                 std::memcpy(dest_buffer, src_ptr, copy_amount);
             },
-            [&system = system, &dest_buffer](const VAddr current_vaddr,
-                                             const std::size_t copy_amount,
-                                             const u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount,
+                const u8* const host_ptr) {
                 if constexpr (!UNSAFE) {
                     system.GPU().FlushRegion(current_vaddr, copy_amount);
                 }
                 std::memcpy(dest_buffer, host_ptr, copy_amount);
             },
-            [&dest_buffer](const std::size_t copy_amount) {
+            [&](const std::size_t copy_amount) {
                 dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
             });
     }
@@ -267,17 +266,16 @@ struct Memory::Impl {
                           "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, dest_addr, size);
             },
-            [&src_buffer](const std::size_t copy_amount, u8* const dest_ptr) {
+            [&](const std::size_t copy_amount, u8* const dest_ptr) {
                 std::memcpy(dest_ptr, src_buffer, copy_amount);
             },
-            [&system = system, &src_buffer](const VAddr current_vaddr,
-                                            const std::size_t copy_amount, u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
                 if constexpr (!UNSAFE) {
                     system.GPU().InvalidateRegion(current_vaddr, copy_amount);
                 }
                 std::memcpy(host_ptr, src_buffer, copy_amount);
             },
-            [&src_buffer](const std::size_t copy_amount) {
+            [&](const std::size_t copy_amount) {
                 src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
             });
     }
@@ -301,8 +299,7 @@ struct Memory::Impl {
             [](const std::size_t copy_amount, u8* const dest_ptr) {
                 std::memset(dest_ptr, 0, copy_amount);
             },
-            [&system = system](const VAddr current_vaddr, const std::size_t copy_amount,
-                               u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
                 system.GPU().InvalidateRegion(current_vaddr, copy_amount);
                 std::memset(host_ptr, 0, copy_amount);
             },
@@ -313,22 +310,20 @@ struct Memory::Impl {
                    const std::size_t size) {
         WalkBlock(
             process, dest_addr, size,
-            [this, &process, &dest_addr, &src_addr, size](const std::size_t copy_amount,
-                                                          const VAddr current_vaddr) {
+            [&](const std::size_t copy_amount, const VAddr current_vaddr) {
                 LOG_ERROR(HW_Memory,
                           "Unmapped CopyBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, src_addr, size);
                 ZeroBlock(process, dest_addr, copy_amount);
             },
-            [this, &process, &dest_addr](const std::size_t copy_amount, const u8* const src_ptr) {
+            [&](const std::size_t copy_amount, const u8* const src_ptr) {
                 WriteBlockImpl<false>(process, dest_addr, src_ptr, copy_amount);
             },
-            [this, &system = system, &process, &dest_addr](
-                const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
                 system.GPU().FlushRegion(current_vaddr, copy_amount);
                 WriteBlockImpl<false>(process, dest_addr, host_ptr, copy_amount);
             },
-            [&dest_addr, &src_addr](const std::size_t copy_amount) {
+            [&](const std::size_t copy_amount) {
                 dest_addr += static_cast<VAddr>(copy_amount);
                 src_addr += static_cast<VAddr>(copy_amount);
             });
@@ -575,7 +570,7 @@ struct Memory::Impl {
             [vaddr]() {
                 LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8, vaddr);
             },
-            [&system = system, vaddr]() { system.GPU().FlushRegion(vaddr, sizeof(T)); });
+            [&]() { system.GPU().FlushRegion(vaddr, sizeof(T)); });
         if (ptr) {
             std::memcpy(&result, ptr, sizeof(T));
         }
@@ -599,7 +594,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped Write{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8,
                           vaddr, static_cast<u64>(data));
             },
-            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
+            [&]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
         if (ptr) {
             std::memcpy(ptr, &data, sizeof(T));
         }
@@ -613,7 +608,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped WriteExclusive{} @ 0x{:016X} = 0x{:016X}",
                           sizeof(T) * 8, vaddr, static_cast<u64>(data));
             },
-            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
+            [&]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
         if (ptr) {
             const auto volatile_pointer = reinterpret_cast<volatile T*>(ptr);
             return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);
@@ -628,7 +623,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped WriteExclusive128 @ 0x{:016X} = 0x{:016X}{:016X}",
                           vaddr, static_cast<u64>(data[1]), static_cast<u64>(data[0]));
             },
-            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(u128)); });
+            [&]() { system.GPU().InvalidateRegion(vaddr, sizeof(u128)); });
         if (ptr) {
             const auto volatile_pointer = reinterpret_cast<volatile u64*>(ptr);
             return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);

src/input_common/CMakeLists.txt

@@ -39,21 +39,14 @@ add_library(input_common STATIC
 if (MSVC)
     target_compile_options(input_common PRIVATE
         /W4
-        /WX
 
         /we4242 # 'identifier': conversion from 'type1' to 'type2', possible loss of data
-        /we4244 # 'conversion': conversion from 'type1' to 'type2', possible loss of data
-        /we4245 # 'conversion': conversion from 'type1' to 'type2', signed/unsigned mismatch
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
+        /we4800 # Implicit conversion from 'type' to bool. Possible information loss
     )
 else()
     target_compile_options(input_common PRIVATE
-        -Werror
         -Werror=conversion
-        -Werror=ignored-qualifiers
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
-        -Werror=unused-variable
     )
 endif()
 

src/input_common/drivers/gc_adapter.cpp

@@ -324,7 +324,7 @@ bool GCAdapter::GetGCEndpoint(libusb_device* device) {
     return true;
 }
 
-Common::Input::VibrationError GCAdapter::SetRumble(
+Common::Input::VibrationError GCAdapter::SetVibration(
     const PadIdentifier& identifier, const Common::Input::VibrationStatus& vibration) {
     const auto mean_amplitude = (vibration.low_amplitude + vibration.high_amplitude) * 0.5f;
     const auto processed_amplitude =
@@ -338,6 +338,10 @@ Common::Input::VibrationError GCAdapter::SetRumble(
     return Common::Input::VibrationError::None;
 }
 
+bool GCAdapter::IsVibrationEnabled([[maybe_unused]] const PadIdentifier& identifier) {
+    return rumble_enabled;
+}
+
 void GCAdapter::UpdateVibrations() {
     // Use 8 states to keep the switching between on/off fast enough for
     // a human to feel different vibration strenght

src/input_common/drivers/gc_adapter.h

@@ -25,9 +25,11 @@ public:
     explicit GCAdapter(std::string input_engine_);
     ~GCAdapter() override;
 
-    Common::Input::VibrationError SetRumble(
+    Common::Input::VibrationError SetVibration(
         const PadIdentifier& identifier, const Common::Input::VibrationStatus& vibration) override;
 
+    bool IsVibrationEnabled(const PadIdentifier& identifier) override;
+
     /// Used for automapping features
     std::vector<Common::ParamPackage> GetInputDevices() const override;
     ButtonMapping GetButtonMappingForDevice(const Common::ParamPackage& params) override;

src/input_common/drivers/sdl_driver.cpp

@@ -40,8 +40,8 @@ public:
     void EnableMotion() {
         if (sdl_controller) {
             SDL_GameController* controller = sdl_controller.get();
-            has_accel = SDL_GameControllerHasSensor(controller, SDL_SENSOR_ACCEL);
-            has_gyro = SDL_GameControllerHasSensor(controller, SDL_SENSOR_GYRO);
+            has_accel = SDL_GameControllerHasSensor(controller, SDL_SENSOR_ACCEL) == SDL_TRUE;
+            has_gyro = SDL_GameControllerHasSensor(controller, SDL_SENSOR_GYRO) == SDL_TRUE;
             if (has_accel) {
                 SDL_GameControllerSetSensorEnabled(controller, SDL_SENSOR_ACCEL, SDL_TRUE);
             }
@@ -114,6 +114,20 @@ public:
         }
         return false;
     }
+
+    void EnableVibration(bool is_enabled) {
+        has_vibration = is_enabled;
+        is_vibration_tested = true;
+    }
+
+    bool HasVibration() const {
+        return has_vibration;
+    }
+
+    bool IsVibrationTested() const {
+        return is_vibration_tested;
+    }
+
     /**
      * The Pad identifier of the joystick
      */
@@ -236,6 +250,8 @@ private:
     u64 last_motion_update{};
     bool has_gyro{false};
     bool has_accel{false};
+    bool has_vibration{false};
+    bool is_vibration_tested{false};
     BasicMotion motion;
 };
 
@@ -517,7 +533,7 @@ std::vector<Common::ParamPackage> SDLDriver::GetInputDevices() const {
     return devices;
 }
 
-Common::Input::VibrationError SDLDriver::SetRumble(
+Common::Input::VibrationError SDLDriver::SetVibration(
     const PadIdentifier& identifier, const Common::Input::VibrationStatus& vibration) {
     const auto joystick =
         GetSDLJoystickByGUID(identifier.guid.RawString(), static_cast<int>(identifier.port));
@@ -546,13 +562,6 @@ Common::Input::VibrationError SDLDriver::SetRumble(
         .type = Common::Input::VibrationAmplificationType::Exponential,
     };
 
-    if (vibration.type == Common::Input::VibrationAmplificationType::Test) {
-        if (!joystick->RumblePlay(new_vibration)) {
-            return Common::Input::VibrationError::Unknown;
-        }
-        return Common::Input::VibrationError::None;
-    }
-
     vibration_queue.Push(VibrationRequest{
         .identifier = identifier,
         .vibration = new_vibration,
@@ -561,6 +570,45 @@ Common::Input::VibrationError SDLDriver::SetRumble(
     return Common::Input::VibrationError::None;
 }
 
+bool SDLDriver::IsVibrationEnabled(const PadIdentifier& identifier) {
+    const auto joystick =
+        GetSDLJoystickByGUID(identifier.guid.RawString(), static_cast<int>(identifier.port));
+
+    constexpr Common::Input::VibrationStatus test_vibration{
+        .low_amplitude = 1,
+        .low_frequency = 160.0f,
+        .high_amplitude = 1,
+        .high_frequency = 320.0f,
+        .type = Common::Input::VibrationAmplificationType::Exponential,
+    };
+
+    constexpr Common::Input::VibrationStatus zero_vibration{
+        .low_amplitude = 0,
+        .low_frequency = 160.0f,
+        .high_amplitude = 0,
+        .high_frequency = 320.0f,
+        .type = Common::Input::VibrationAmplificationType::Exponential,
+    };
+
+    if (joystick->IsVibrationTested()) {
+        return joystick->HasVibration();
+    }
+
+    // First vibration might fail
+    joystick->RumblePlay(test_vibration);
+
+    // Wait for about 15ms to ensure the controller is ready for the stop command
+    std::this_thread::sleep_for(std::chrono::milliseconds(15));
+
+    if (!joystick->RumblePlay(zero_vibration)) {
+        joystick->EnableVibration(false);
+        return false;
+    }
+
+    joystick->EnableVibration(true);
+    return true;
+}
+
 void SDLDriver::SendVibrations() {
     while (!vibration_queue.Empty()) {
         VibrationRequest request;

src/input_common/drivers/sdl_driver.h

@@ -61,9 +61,11 @@ public:
 
     bool IsStickInverted(const Common::ParamPackage& params) override;
 
-    Common::Input::VibrationError SetRumble(
+    Common::Input::VibrationError SetVibration(
         const PadIdentifier& identifier, const Common::Input::VibrationStatus& vibration) override;
 
+    bool IsVibrationEnabled(const PadIdentifier& identifier) override;
+
 private:
     struct VibrationRequest {
         PadIdentifier identifier;

src/input_common/input_engine.h

@@ -108,12 +108,17 @@ public:
                          [[maybe_unused]] const Common::Input::LedStatus& led_status) {}
 
     // Sets rumble to a controller
-    virtual Common::Input::VibrationError SetRumble(
+    virtual Common::Input::VibrationError SetVibration(
         [[maybe_unused]] const PadIdentifier& identifier,
         [[maybe_unused]] const Common::Input::VibrationStatus& vibration) {
         return Common::Input::VibrationError::NotSupported;
     }
 
+    // Returns true if device supports vibrations
+    virtual bool IsVibrationEnabled([[maybe_unused]] const PadIdentifier& identifier) {
+        return false;
+    }
+
     // Sets polling mode to a controller
     virtual Common::Input::PollingError SetPollingMode(
         [[maybe_unused]] const PadIdentifier& identifier,

src/input_common/input_poller.cpp

@@ -763,7 +763,11 @@ public:
 
     Common::Input::VibrationError SetVibration(
         const Common::Input::VibrationStatus& vibration_status) override {
-        return input_engine->SetRumble(identifier, vibration_status);
+        return input_engine->SetVibration(identifier, vibration_status);
+    }
+
+    bool IsVibrationEnabled() override {
+        return input_engine->IsVibrationEnabled(identifier);
     }
 
     Common::Input::PollingError SetPollingMode(Common::Input::PollingMode polling_mode) override {
@@ -797,8 +801,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateButtonDevice(
 
     const auto button_id = params.Get("button", 0);
     const auto keyboard_key = params.Get("code", 0);
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
     input_engine->PreSetController(identifier);
     input_engine->PreSetButton(identifier, button_id);
     input_engine->PreSetButton(identifier, keyboard_key);
@@ -820,8 +824,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateHatButtonDevice(
 
     const auto button_id = params.Get("hat", 0);
     const auto direction = input_engine->GetHatButtonId(params.Get("direction", ""));
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
 
     input_engine->PreSetController(identifier);
     input_engine->PreSetHatButton(identifier, button_id);
@@ -879,7 +883,7 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateAnalogDevice(
         .threshold = std::clamp(params.Get("threshold", 0.5f), 0.0f, 1.0f),
         .offset = std::clamp(params.Get("offset", 0.0f), -1.0f, 1.0f),
         .inverted = params.Get("invert", "+") == "-",
-        .toggle = static_cast<bool>(params.Get("toggle", false)),
+        .toggle = params.Get("toggle", false) != 0,
     };
     input_engine->PreSetController(identifier);
     input_engine->PreSetAxis(identifier, axis);
@@ -895,8 +899,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateTriggerDevice(
     };
 
     const auto button = params.Get("button", 0);
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
 
     const auto axis = params.Get("axis", 0);
     const Common::Input::AnalogProperties properties = {
@@ -926,8 +930,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateTouchDevice(
     };
 
     const auto button = params.Get("button", 0);
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
 
     const auto axis_x = params.Get("axis_x", 0);
     const Common::Input::AnalogProperties properties_x = {

src/shader_recompiler/CMakeLists.txt

@@ -241,24 +241,14 @@ target_link_libraries(shader_recompiler PUBLIC common fmt::fmt sirit)
 if (MSVC)
     target_compile_options(shader_recompiler PRIVATE
         /W4
-        /WX
-        /we4018 # 'expression' : signed/unsigned mismatch
-        /we4244 # 'argument' : conversion from 'type1' to 'type2', possible loss of data (floating-point)
-        /we4245 # 'conversion' : conversion from 'type1' to 'type2', signed/unsigned mismatch
+
+        /we4242 # 'identifier': conversion from 'type1' to 'type2', possible loss of data
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
-        /we4267 # 'var' : conversion from 'size_t' to 'type', possible loss of data
-        /we4305 # 'context' : truncation from 'type1' to 'type2'
         /we4800 # Implicit conversion from 'type' to bool. Possible information loss
-        /we4826 # Conversion from 'type1' to 'type2' is sign-extended. This may cause unexpected runtime behavior.
     )
 else()
     target_compile_options(shader_recompiler PRIVATE
-        -Werror
         -Werror=conversion
-        -Werror=ignored-qualifiers
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
-        -Werror=unused-variable
 
         # Bracket depth determines maximum size of a fold expression in Clang since 9c9974c3ccb6.
         # And this in turns limits the size of a std::array.

src/shader_recompiler/backend/glasm/emit_glasm_not_implemented.cpp

@@ -5,10 +5,6 @@
 #include "shader_recompiler/backend/glasm/glasm_emit_context.h"
 #include "shader_recompiler/frontend/ir/value.h"
 
-#ifdef _MSC_VER
-#pragma warning(disable : 4100)
-#endif
-
 namespace Shader::Backend::GLASM {
 
 #define NotImplemented() throw NotImplementedException("GLASM instruction {}", __LINE__)

src/shader_recompiler/backend/glsl/emit_glsl_not_implemented.cpp

@@ -7,10 +7,6 @@
 #include "shader_recompiler/backend/glsl/glsl_emit_context.h"
 #include "shader_recompiler/frontend/ir/value.h"
 
-#ifdef _MSC_VER
-#pragma warning(disable : 4100)
-#endif
-
 namespace Shader::Backend::GLSL {
 
 void EmitGetRegister(EmitContext& ctx) {

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -137,28 +137,35 @@ bool IsLegacyAttribute(IR::Attribute attribute) {
 }
 
 std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
-    const VaryingState& state, std::queue<IR::Attribute> ununsed_generics) {
+    const VaryingState& state, std::queue<IR::Attribute> unused_generics,
+    const std::map<IR::Attribute, IR::Attribute>& previous_stage_mapping) {
     std::map<IR::Attribute, IR::Attribute> mapping;
+    auto update_mapping = [&mapping, &unused_generics, previous_stage_mapping](IR::Attribute attr,
+                                                                               size_t count) {
+        if (previous_stage_mapping.find(attr) != previous_stage_mapping.end()) {
+            for (size_t i = 0; i < count; ++i) {
+                mapping.insert({attr + i, previous_stage_mapping.at(attr + i)});
+            }
+        } else {
+            for (size_t i = 0; i < count; ++i) {
+                mapping.insert({attr + i, unused_generics.front() + i});
+            }
+            unused_generics.pop();
+        }
+    };
     for (size_t index = 0; index < 4; ++index) {
         auto attr = IR::Attribute::ColorFrontDiffuseR + index * 4;
         if (state.AnyComponent(attr)) {
-            for (size_t i = 0; i < 4; ++i) {
-                mapping.insert({attr + i, ununsed_generics.front() + i});
-            }
-            ununsed_generics.pop();
+            update_mapping(attr, 4);
         }
     }
     if (state[IR::Attribute::FogCoordinate]) {
-        mapping.insert({IR::Attribute::FogCoordinate, ununsed_generics.front()});
-        ununsed_generics.pop();
+        update_mapping(IR::Attribute::FogCoordinate, 1);
     }
     for (size_t index = 0; index < IR::NUM_FIXEDFNCTEXTURE; ++index) {
         auto attr = IR::Attribute::FixedFncTexture0S + index * 4;
         if (state.AnyComponent(attr)) {
-            for (size_t i = 0; i < 4; ++i) {
-                mapping.insert({attr + i, ununsed_generics.front() + i});
-            }
-            ununsed_generics.pop();
+            update_mapping(attr, 4);
         }
     }
     return mapping;
@@ -265,21 +272,22 @@ IR::Program MergeDualVertexPrograms(IR::Program& vertex_a, IR::Program& vertex_b
 void ConvertLegacyToGeneric(IR::Program& program, const Shader::RuntimeInfo& runtime_info) {
     auto& stores = program.info.stores;
     if (stores.Legacy()) {
-        std::queue<IR::Attribute> ununsed_output_generics{};
+        std::queue<IR::Attribute> unused_output_generics{};
         for (size_t index = 0; index < IR::NUM_GENERICS; ++index) {
             if (!stores.Generic(index)) {
-                ununsed_output_generics.push(IR::Attribute::Generic0X + index * 4);
+                unused_output_generics.push(IR::Attribute::Generic0X + index * 4);
             }
         }
-        auto mappings = GenerateLegacyToGenericMappings(stores, ununsed_output_generics);
+        program.info.legacy_stores_mapping =
+            GenerateLegacyToGenericMappings(stores, unused_output_generics, {});
         for (IR::Block* const block : program.post_order_blocks) {
             for (IR::Inst& inst : block->Instructions()) {
                 switch (inst.GetOpcode()) {
                 case IR::Opcode::SetAttribute: {
                     const auto attr = inst.Arg(0).Attribute();
                     if (IsLegacyAttribute(attr)) {
-                        stores.Set(mappings[attr], true);
-                        inst.SetArg(0, Shader::IR::Value(mappings[attr]));
+                        stores.Set(program.info.legacy_stores_mapping[attr], true);
+                        inst.SetArg(0, Shader::IR::Value(program.info.legacy_stores_mapping[attr]));
                     }
                     break;
                 }
@@ -292,15 +300,16 @@ void ConvertLegacyToGeneric(IR::Program& program, const Shader::RuntimeInfo& run
 
     auto& loads = program.info.loads;
     if (loads.Legacy()) {
-        std::queue<IR::Attribute> ununsed_input_generics{};
+        std::queue<IR::Attribute> unused_input_generics{};
         for (size_t index = 0; index < IR::NUM_GENERICS; ++index) {
             const AttributeType input_type{runtime_info.generic_input_types[index]};
             if (!runtime_info.previous_stage_stores.Generic(index) || !loads.Generic(index) ||
                 input_type == AttributeType::Disabled) {
-                ununsed_input_generics.push(IR::Attribute::Generic0X + index * 4);
+                unused_input_generics.push(IR::Attribute::Generic0X + index * 4);
             }
         }
-        auto mappings = GenerateLegacyToGenericMappings(loads, ununsed_input_generics);
+        auto mappings = GenerateLegacyToGenericMappings(
+            loads, unused_input_generics, runtime_info.previous_stage_legacy_stores_mapping);
         for (IR::Block* const block : program.post_order_blocks) {
             for (IR::Inst& inst : block->Instructions()) {
                 switch (inst.GetOpcode()) {

src/shader_recompiler/runtime_info.h

@@ -4,6 +4,7 @@
 #pragma once
 
 #include <array>
+#include <map>
 #include <optional>
 #include <vector>
 
@@ -60,6 +61,7 @@ struct TransformFeedbackVarying {
 struct RuntimeInfo {
     std::array<AttributeType, 32> generic_input_types{};
     VaryingState previous_stage_stores;
+    std::map<IR::Attribute, IR::Attribute> previous_stage_legacy_stores_mapping;
 
     bool convert_depth_mode{};
     bool force_early_z{};

src/shader_recompiler/shader_info.h

@@ -5,6 +5,7 @@
 
 #include <array>
 #include <bitset>
+#include <map>
 
 #include "common/common_types.h"
 #include "shader_recompiler/frontend/ir/type.h"
@@ -127,6 +128,8 @@ struct Info {
     VaryingState stores;
     VaryingState passthrough;
 
+    std::map<IR::Attribute, IR::Attribute> legacy_stores_mapping;
+
     bool loads_indexed_attributes{};
 
     std::array<bool, 8> stores_frag_color{};

src/tests/video_core/buffer_base.cpp

@@ -44,7 +44,7 @@ public:
 
     [[nodiscard]] unsigned Count() const noexcept {
         unsigned count = 0;
-        for (const auto [index, value] : page_table) {
+        for (const auto& [index, value] : page_table) {
             count += value;
         }
         return count;

src/video_core/CMakeLists.txt

@@ -279,14 +279,8 @@ if (MSVC)
 else()
     target_compile_options(video_core PRIVATE
         -Werror=conversion
-        -Wno-error=sign-conversion
-        -Werror=pessimizing-move
-        -Werror=redundant-move
-        -Werror=type-limits
 
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=class-memaccess>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
+        -Wno-sign-conversion
     )
 endif()
 

src/video_core/engines/maxwell_3d.cpp

@@ -117,10 +117,18 @@ void Maxwell3D::InitializeRegisterDefaults() {
 
     shadow_state = regs;
 
-    mme_inline[MAXWELL3D_REG_INDEX(draw.end)] = true;
-    mme_inline[MAXWELL3D_REG_INDEX(draw.begin)] = true;
-    mme_inline[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
-    mme_inline[MAXWELL3D_REG_INDEX(index_buffer.count)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(draw.end)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(draw.begin)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(vertex_buffer.first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer.first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer.count)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer32_first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer16_first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(index_buffer8_first)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(draw_inline_index)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(inline_index_2x16.even)] = true;
+    draw_command[MAXWELL3D_REG_INDEX(inline_index_4x8.index0)] = true;
 }
 
 void Maxwell3D::ProcessMacro(u32 method, const u32* base_start, u32 amount, bool is_last_call) {
@@ -208,25 +216,6 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume
         return ProcessCBBind(3);
     case MAXWELL3D_REG_INDEX(bind_groups[4].raw_config):
         return ProcessCBBind(4);
-    case MAXWELL3D_REG_INDEX(draw.end):
-        return DrawArrays();
-    case MAXWELL3D_REG_INDEX(index_buffer32_first):
-        regs.index_buffer.count = regs.index_buffer32_first.count;
-        regs.index_buffer.first = regs.index_buffer32_first.first;
-        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        return DrawArrays();
-    case MAXWELL3D_REG_INDEX(index_buffer16_first):
-        regs.index_buffer.count = regs.index_buffer16_first.count;
-        regs.index_buffer.first = regs.index_buffer16_first.first;
-        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        return DrawArrays();
-    case MAXWELL3D_REG_INDEX(index_buffer8_first):
-        regs.index_buffer.count = regs.index_buffer8_first.count;
-        regs.index_buffer.first = regs.index_buffer8_first.first;
-        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        // a macro calls this one over and over, should it increase instancing?
-        // Used by Hades and likely other Vulkan games.
-        return DrawArrays();
     case MAXWELL3D_REG_INDEX(topology_override):
         use_topology_override = true;
         return;
@@ -261,14 +250,13 @@ void Maxwell3D::CallMacroMethod(u32 method, const std::vector<u32>& parameters)
 
     // Execute the current macro.
     macro_engine->Execute(macro_positions[entry], parameters);
-    if (mme_draw.current_mode != MMEDrawMode::Undefined) {
-        FlushMMEInlineDraw();
-    }
+
+    ProcessDeferredDraw();
 }
 
 void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
-    // It is an error to write to a register other than the current macro's ARG register before it
-    // has finished execution.
+    // It is an error to write to a register other than the current macro's ARG register before
+    // it has finished execution.
     if (executing_macro != 0) {
         ASSERT(method == executing_macro + 1);
     }
@@ -283,9 +271,33 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid Maxwell3D register, increase the size of the Regs structure");
 
-    const u32 argument = ProcessShadowRam(method, method_argument);
-    ProcessDirtyRegisters(method, argument);
-    ProcessMethodCall(method, argument, method_argument, is_last_call);
+    if (draw_command[method]) {
+        regs.reg_array[method] = method_argument;
+        deferred_draw_method.push_back(method);
+        auto u32_to_u8 = [&](const u32 argument) {
+            inline_index_draw_indexes.push_back(static_cast<u8>(argument & 0x000000ff));
+            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0x0000ff00) >> 8));
+            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0x00ff0000) >> 16));
+            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0xff000000) >> 24));
+        };
+        if (MAXWELL3D_REG_INDEX(draw_inline_index) == method) {
+            u32_to_u8(method_argument);
+        } else if (MAXWELL3D_REG_INDEX(inline_index_2x16.even) == method) {
+            u32_to_u8(regs.inline_index_2x16.even);
+            u32_to_u8(regs.inline_index_2x16.odd);
+        } else if (MAXWELL3D_REG_INDEX(inline_index_4x8.index0) == method) {
+            u32_to_u8(regs.inline_index_4x8.index0);
+            u32_to_u8(regs.inline_index_4x8.index1);
+            u32_to_u8(regs.inline_index_4x8.index2);
+            u32_to_u8(regs.inline_index_4x8.index3);
+        }
+    } else {
+        ProcessDeferredDraw();
+
+        const u32 argument = ProcessShadowRam(method, method_argument);
+        ProcessDirtyRegisters(method, argument);
+        ProcessMethodCall(method, argument, method_argument, is_last_call);
+    }
 }
 
 void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
@@ -326,55 +338,6 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
     }
 }
 
-void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) {
-    if (mme_draw.current_mode == MMEDrawMode::Undefined) {
-        if (mme_draw.gl_begin_consume) {
-            mme_draw.current_mode = expected_mode;
-            mme_draw.current_count = count;
-            mme_draw.instance_count = 1;
-            mme_draw.gl_begin_consume = false;
-            mme_draw.gl_end_count = 0;
-        }
-        return;
-    } else {
-        if (mme_draw.current_mode == expected_mode && count == mme_draw.current_count &&
-            mme_draw.instance_mode && mme_draw.gl_begin_consume) {
-            mme_draw.instance_count++;
-            mme_draw.gl_begin_consume = false;
-            return;
-        } else {
-            FlushMMEInlineDraw();
-        }
-    }
-    // Tail call in case it needs to retry.
-    StepInstance(expected_mode, count);
-}
-
-void Maxwell3D::CallMethodFromMME(u32 method, u32 method_argument) {
-    if (mme_inline[method]) {
-        regs.reg_array[method] = method_argument;
-        if (method == MAXWELL3D_REG_INDEX(vertex_buffer.count) ||
-            method == MAXWELL3D_REG_INDEX(index_buffer.count)) {
-            const MMEDrawMode expected_mode = method == MAXWELL3D_REG_INDEX(vertex_buffer.count)
-                                                  ? MMEDrawMode::Array
-                                                  : MMEDrawMode::Indexed;
-            StepInstance(expected_mode, method_argument);
-        } else if (method == MAXWELL3D_REG_INDEX(draw.begin)) {
-            mme_draw.instance_mode =
-                (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) ||
-                (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Unchanged);
-            mme_draw.gl_begin_consume = true;
-        } else {
-            mme_draw.gl_end_count++;
-        }
-    } else {
-        if (mme_draw.current_mode != MMEDrawMode::Undefined) {
-            FlushMMEInlineDraw();
-        }
-        CallMethod(method, method_argument, true);
-    }
-}
-
 void Maxwell3D::ProcessTopologyOverride() {
     using PrimitiveTopology = Maxwell3D::Regs::PrimitiveTopology;
     using PrimitiveTopologyOverride = Maxwell3D::Regs::PrimitiveTopologyOverride;
@@ -404,41 +367,6 @@ void Maxwell3D::ProcessTopologyOverride() {
     }
 }
 
-void Maxwell3D::FlushMMEInlineDraw() {
-    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
-              regs.vertex_buffer.count);
-    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
-    ASSERT(mme_draw.instance_count == mme_draw.gl_end_count);
-
-    // Both instance configuration registers can not be set at the same time.
-    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
-                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
-               "Illegal combination of instancing parameters");
-
-    ProcessTopologyOverride();
-
-    const bool is_indexed = mme_draw.current_mode == MMEDrawMode::Indexed;
-    if (ShouldExecute()) {
-        rasterizer->Draw(is_indexed, true);
-    }
-
-    // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
-    // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
-    // it's possible that it is incorrect and that there is some other register used to specify the
-    // drawing mode.
-    if (is_indexed) {
-        regs.index_buffer.count = 0;
-    } else {
-        regs.vertex_buffer.count = 0;
-    }
-    mme_draw.current_mode = MMEDrawMode::Undefined;
-    mme_draw.current_count = 0;
-    mme_draw.instance_count = 0;
-    mme_draw.instance_mode = false;
-    mme_draw.gl_begin_consume = false;
-    mme_draw.gl_end_count = 0;
-}
-
 void Maxwell3D::ProcessMacroUpload(u32 data) {
     macro_engine->AddCode(regs.load_mme.instruction_ptr++, data);
 }
@@ -573,42 +501,6 @@ void Maxwell3D::ProcessSyncPoint() {
     rasterizer->SignalSyncPoint(sync_point);
 }
 
-void Maxwell3D::DrawArrays() {
-    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
-              regs.vertex_buffer.count);
-    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
-
-    // Both instance configuration registers can not be set at the same time.
-    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
-                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
-               "Illegal combination of instancing parameters");
-
-    ProcessTopologyOverride();
-
-    if (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) {
-        // Increment the current instance *before* drawing.
-        state.current_instance++;
-    } else if (regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged) {
-        // Reset the current instance to 0.
-        state.current_instance = 0;
-    }
-
-    const bool is_indexed{regs.index_buffer.count && !regs.vertex_buffer.count};
-    if (ShouldExecute()) {
-        rasterizer->Draw(is_indexed, false);
-    }
-
-    // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
-    // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
-    // it's possible that it is incorrect and that there is some other register used to specify the
-    // drawing mode.
-    if (is_indexed) {
-        regs.index_buffer.count = 0;
-    } else {
-        regs.vertex_buffer.count = 0;
-    }
-}
-
 std::optional<u64> Maxwell3D::GetQueryResult() {
     switch (regs.report_semaphore.query.report) {
     case Regs::ReportSemaphore::Report::Payload:
@@ -691,4 +583,95 @@ void Maxwell3D::ProcessClearBuffers() {
     rasterizer->Clear();
 }
 
+void Maxwell3D::ProcessDeferredDraw() {
+    if (deferred_draw_method.empty()) {
+        return;
+    }
+
+    enum class DrawMode {
+        Undefined,
+        General,
+        Instance,
+    };
+    DrawMode draw_mode{DrawMode::Undefined};
+    u32 instance_count = 1;
+
+    auto first_method = deferred_draw_method[0];
+    if (MAXWELL3D_REG_INDEX(draw.begin) == first_method) {
+        // The minimum number of methods for drawing must be greater than or equal to
+        // 3[draw.begin->vertex(index)count->draw.end] to avoid errors in index mode drawing
+        if (deferred_draw_method.size() < 3) {
+            return;
+        }
+        draw_mode = (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) ||
+                            (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Unchanged)
+                        ? DrawMode::Instance
+                        : DrawMode::General;
+    } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method ||
+               MAXWELL3D_REG_INDEX(index_buffer16_first) == first_method ||
+               MAXWELL3D_REG_INDEX(index_buffer8_first) == first_method) {
+        draw_mode = DrawMode::General;
+    }
+
+    // Drawing will only begin with draw.begin or index_buffer method, other methods directly
+    // clear
+    if (draw_mode == DrawMode::Undefined) {
+        deferred_draw_method.clear();
+        return;
+    }
+
+    if (draw_mode == DrawMode::Instance) {
+        ASSERT_MSG(deferred_draw_method.size() % 4 == 0, "Instance mode method size error");
+        instance_count = static_cast<u32>(deferred_draw_method.size()) / 4;
+    } else {
+        if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
+            regs.index_buffer.count = regs.index_buffer32_first.count;
+            regs.index_buffer.first = regs.index_buffer32_first.first;
+            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
+            regs.index_buffer.count = regs.index_buffer16_first.count;
+            regs.index_buffer.first = regs.index_buffer16_first.first;
+            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
+            regs.index_buffer.count = regs.index_buffer8_first.count;
+            regs.index_buffer.first = regs.index_buffer8_first.first;
+            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        } else {
+            auto second_method = deferred_draw_method[1];
+            if (MAXWELL3D_REG_INDEX(draw_inline_index) == second_method ||
+                MAXWELL3D_REG_INDEX(inline_index_2x16.even) == second_method ||
+                MAXWELL3D_REG_INDEX(inline_index_4x8.index0) == second_method) {
+                regs.index_buffer.count = static_cast<u32>(inline_index_draw_indexes.size() / 4);
+                regs.index_buffer.format = Regs::IndexFormat::UnsignedInt;
+            }
+        }
+    }
+
+    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
+              regs.vertex_buffer.count);
+
+    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
+
+    // Both instance configuration registers can not be set at the same time.
+    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
+                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
+               "Illegal combination of instancing parameters");
+
+    ProcessTopologyOverride();
+
+    const bool is_indexed = regs.index_buffer.count && !regs.vertex_buffer.count;
+    if (ShouldExecute()) {
+        rasterizer->Draw(is_indexed, instance_count);
+    }
+
+    if (is_indexed) {
+        regs.index_buffer.count = 0;
+    } else {
+        regs.vertex_buffer.count = 0;
+    }
+
+    deferred_draw_method.clear();
+    inline_index_draw_indexes.clear();
+}
+
 } // namespace Tegra::Engines

src/video_core/engines/maxwell_3d.h

@@ -1739,14 +1739,11 @@ public:
             Footprint_1x1_Virtual = 2,
         };
 
-        struct InlineIndex4x8Align {
+        struct InlineIndex4x8 {
             union {
                 BitField<0, 30, u32> count;
                 BitField<30, 2, u32> start;
             };
-        };
-
-        struct InlineIndex4x8Index {
             union {
                 BitField<0, 8, u32> index0;
                 BitField<8, 8, u32> index1;
@@ -2836,8 +2833,7 @@ public:
                 u32 depth_write_enabled;                                               ///< 0x12E8
                 u32 alpha_test_enabled;                                                ///< 0x12EC
                 INSERT_PADDING_BYTES_NOINIT(0x10);
-                InlineIndex4x8Align inline_index_4x8_align;                            ///< 0x1300
-                InlineIndex4x8Index inline_index_4x8_index;                            ///< 0x1304
+                InlineIndex4x8 inline_index_4x8;                                       ///< 0x1300
                 D3DCullMode d3d_cull_mode;                                             ///< 0x1308
                 ComparisonOp depth_test_func;                                          ///< 0x130C
                 f32 alpha_test_ref;                                                    ///< 0x1310
@@ -3048,8 +3044,6 @@ public:
         };
 
         std::array<ShaderStageInfo, Regs::MaxShaderStage> shader_stages;
-
-        u32 current_instance = 0; ///< Current instance to be used to simulate instanced rendering.
     };
 
     State state{};
@@ -3064,11 +3058,6 @@ public:
     void CallMultiMethod(u32 method, const u32* base_start, u32 amount,
                          u32 methods_pending) override;
 
-    /// Write the value to the register identified by method.
-    void CallMethodFromMME(u32 method, u32 method_argument);
-
-    void FlushMMEInlineDraw();
-
     bool ShouldExecute() const {
         return execute_on;
     }
@@ -3081,21 +3070,6 @@ public:
         return *rasterizer;
     }
 
-    enum class MMEDrawMode : u32 {
-        Undefined,
-        Array,
-        Indexed,
-    };
-
-    struct MMEDrawState {
-        MMEDrawMode current_mode{MMEDrawMode::Undefined};
-        u32 current_count{};
-        u32 instance_count{};
-        bool instance_mode{};
-        bool gl_begin_consume{};
-        u32 gl_end_count{};
-    } mme_draw;
-
     struct DirtyState {
         using Flags = std::bitset<std::numeric_limits<u8>::max()>;
         using Table = std::array<u8, Regs::NUM_REGS>;
@@ -3105,6 +3079,8 @@ public:
         Tables tables{};
     } dirty;
 
+    std::vector<u8> inline_index_draw_indexes;
+
 private:
     void InitializeRegisterDefaults();
 
@@ -3164,14 +3140,10 @@ private:
     /// Handles a write to the CB_BIND register.
     void ProcessCBBind(size_t stage_index);
 
-    /// Handles a write to the VERTEX_END_GL register, triggering a draw.
-    void DrawArrays();
-
     /// Handles use of topology overrides (e.g., to avoid using a topology assigned from a macro)
     void ProcessTopologyOverride();
 
-    // Handles a instance drawcall from MME
-    void StepInstance(MMEDrawMode expected_mode, u32 count);
+    void ProcessDeferredDraw();
 
     /// Returns a query's value or an empty object if the value will be deferred through a cache.
     std::optional<u64> GetQueryResult();
@@ -3184,8 +3156,6 @@ private:
     /// Start offsets of each macro in macro_memory
     std::array<u32, 0x80> macro_positions{};
 
-    std::array<bool, Regs::NUM_REGS> mme_inline{};
-
     /// Macro method that is currently being executed / being fed parameters.
     u32 executing_macro = 0;
     /// Parameters that have been submitted to the macro call so far.
@@ -3198,6 +3168,9 @@ private:
 
     bool execute_on{true};
     bool use_topology_override{false};
+
+    std::array<bool, Regs::NUM_REGS> draw_command{};
+    std::vector<u32> deferred_draw_method;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
@@ -3402,8 +3375,7 @@ ASSERT_REG_POSITION(alpha_to_coverage_dither, 0x12E0);
 ASSERT_REG_POSITION(blend_per_target_enabled, 0x12E4);
 ASSERT_REG_POSITION(depth_write_enabled, 0x12E8);
 ASSERT_REG_POSITION(alpha_test_enabled, 0x12EC);
-ASSERT_REG_POSITION(inline_index_4x8_align, 0x1300);
-ASSERT_REG_POSITION(inline_index_4x8_index, 0x1304);
+ASSERT_REG_POSITION(inline_index_4x8, 0x1300);
 ASSERT_REG_POSITION(d3d_cull_mode, 0x1308);
 ASSERT_REG_POSITION(depth_test_func, 0x130C);
 ASSERT_REG_POSITION(alpha_test_ref, 0x1310);

src/video_core/macro/macro_hle.cpp

@@ -22,35 +22,29 @@ void HLE_771BB18C62444DA0(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0] & 0x3ffffff));
     maxwell3d.regs.global_base_instance_index = parameters[5];
-    maxwell3d.mme_draw.instance_count = instance_count;
     maxwell3d.regs.global_base_vertex_index = parameters[3];
     maxwell3d.regs.index_buffer.count = parameters[1];
     maxwell3d.regs.index_buffer.first = parameters[4];
 
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(true, true);
+        maxwell3d.Rasterizer().Draw(true, instance_count);
     }
     maxwell3d.regs.index_buffer.count = 0;
-    maxwell3d.mme_draw.instance_count = 0;
-    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
 }
 
 void HLE_0D61FC9FAAC9FCAD(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& parameters) {
-    const u32 count = (maxwell3d.GetRegisterValue(0xD1B) & parameters[2]);
+    const u32 instance_count = (maxwell3d.GetRegisterValue(0xD1B) & parameters[2]);
 
     maxwell3d.regs.vertex_buffer.first = parameters[3];
     maxwell3d.regs.vertex_buffer.count = parameters[1];
     maxwell3d.regs.global_base_instance_index = parameters[4];
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0]));
-    maxwell3d.mme_draw.instance_count = count;
 
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(false, true);
+        maxwell3d.Rasterizer().Draw(false, instance_count);
     }
     maxwell3d.regs.vertex_buffer.count = 0;
-    maxwell3d.mme_draw.instance_count = 0;
-    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
 }
 
 void HLE_0217920100488FF7(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& parameters) {
@@ -63,24 +57,21 @@ void HLE_0217920100488FF7(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
     maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
     maxwell3d.regs.global_base_vertex_index = element_base;
     maxwell3d.regs.global_base_instance_index = base_instance;
-    maxwell3d.mme_draw.instance_count = instance_count;
-    maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-    maxwell3d.CallMethodFromMME(0x8e4, element_base);
-    maxwell3d.CallMethodFromMME(0x8e5, base_instance);
+    maxwell3d.CallMethod(0x8e3, 0x640, true);
+    maxwell3d.CallMethod(0x8e4, element_base, true);
+    maxwell3d.CallMethod(0x8e5, base_instance, true);
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0]));
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(true, true);
+        maxwell3d.Rasterizer().Draw(true, instance_count);
     }
     maxwell3d.regs.vertex_id_base = 0x0;
     maxwell3d.regs.index_buffer.count = 0;
     maxwell3d.regs.global_base_vertex_index = 0x0;
     maxwell3d.regs.global_base_instance_index = 0x0;
-    maxwell3d.mme_draw.instance_count = 0;
-    maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-    maxwell3d.CallMethodFromMME(0x8e4, 0x0);
-    maxwell3d.CallMethodFromMME(0x8e5, 0x0);
-    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
+    maxwell3d.CallMethod(0x8e3, 0x640, true);
+    maxwell3d.CallMethod(0x8e4, 0x0, true);
+    maxwell3d.CallMethod(0x8e5, 0x0, true);
 }
 
 // Multidraw Indirect
@@ -91,11 +82,9 @@ void HLE_3F5E74B9C9A50164(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
         maxwell3d.regs.index_buffer.count = 0;
         maxwell3d.regs.global_base_vertex_index = 0x0;
         maxwell3d.regs.global_base_instance_index = 0x0;
-        maxwell3d.mme_draw.instance_count = 0;
-        maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-        maxwell3d.CallMethodFromMME(0x8e4, 0x0);
-        maxwell3d.CallMethodFromMME(0x8e5, 0x0);
-        maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
+        maxwell3d.CallMethod(0x8e3, 0x640, true);
+        maxwell3d.CallMethod(0x8e4, 0x0, true);
+        maxwell3d.CallMethod(0x8e5, 0x0, true);
         maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
     });
     const u32 start_indirect = parameters[0];
@@ -127,15 +116,13 @@ void HLE_3F5E74B9C9A50164(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
         maxwell3d.regs.index_buffer.count = num_vertices;
         maxwell3d.regs.global_base_vertex_index = base_vertex;
         maxwell3d.regs.global_base_instance_index = base_instance;
-        maxwell3d.mme_draw.instance_count = instance_count;
-        maxwell3d.CallMethodFromMME(0x8e3, 0x640);
-        maxwell3d.CallMethodFromMME(0x8e4, base_vertex);
-        maxwell3d.CallMethodFromMME(0x8e5, base_instance);
+        maxwell3d.CallMethod(0x8e3, 0x640, true);
+        maxwell3d.CallMethod(0x8e4, base_vertex, true);
+        maxwell3d.CallMethod(0x8e5, base_instance, true);
         maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
         if (maxwell3d.ShouldExecute()) {
-            maxwell3d.Rasterizer().Draw(true, true);
+            maxwell3d.Rasterizer().Draw(true, instance_count);
         }
-        maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
     }
 }
 

src/video_core/macro/macro_interpreter.cpp

@@ -335,7 +335,7 @@ void MacroInterpreterImpl::SetMethodAddress(u32 address) {
 }
 
 void MacroInterpreterImpl::Send(u32 value) {
-    maxwell3d.CallMethodFromMME(method_address.address, value);
+    maxwell3d.CallMethod(method_address.address, value, true);
     // Increment the method address by the method increment.
     method_address.address.Assign(method_address.address.Value() +
                                   method_address.increment.Value());

src/video_core/macro/macro_jit_x64.cpp

@@ -346,7 +346,7 @@ void MacroJITx64Impl::Compile_Read(Macro::Opcode opcode) {
 }
 
 void Send(Engines::Maxwell3D* maxwell3d, Macro::MethodAddress method_address, u32 value) {
-    maxwell3d->CallMethodFromMME(method_address.address, value);
+    maxwell3d->CallMethod(method_address.address, value, true);
 }
 
 void MacroJITx64Impl::Compile_Send(Xbyak::Reg32 value) {

src/video_core/memory_manager.cpp

@@ -133,7 +133,7 @@ inline void MemoryManager::SetBigPageContinous(size_t big_page_index, bool value
 template <MemoryManager::EntryType entry_type>
 GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
                                     PTEKind kind) {
-    u64 remaining_size{size};
+    [[maybe_unused]] u64 remaining_size{size};
     if constexpr (entry_type == EntryType::Mapped) {
         page_table.ReserveRange(gpu_addr, size);
     }
@@ -159,7 +159,7 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cp
 template <MemoryManager::EntryType entry_type>
 GPUVAddr MemoryManager::BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr,
                                        size_t size, PTEKind kind) {
-    u64 remaining_size{size};
+    [[maybe_unused]] u64 remaining_size{size};
     for (u64 offset{}; offset < size; offset += big_page_size) {
         const GPUVAddr current_gpu_addr = gpu_addr + offset;
         [[maybe_unused]] const auto current_entry_type = GetEntry<true>(current_gpu_addr);

src/video_core/rasterizer_interface.h

@@ -40,7 +40,7 @@ public:
     virtual ~RasterizerInterface() = default;
 
     /// Dispatches a draw invocation
-    virtual void Draw(bool is_indexed, bool is_instanced) = 0;
+    virtual void Draw(bool is_indexed, u32 instance_count) = 0;
 
     /// Clear the current framebuffer
     virtual void Clear() = 0;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -205,7 +205,7 @@ void RasterizerOpenGL::Clear() {
     ++num_queued_commands;
 }
 
-void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
+void RasterizerOpenGL::Draw(bool is_indexed, u32 instance_count) {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
@@ -222,14 +222,15 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     pipeline->SetEngine(maxwell3d, gpu_memory);
     pipeline->Configure(is_indexed);
 
+    BindInlineIndexBuffer();
+
     SyncState();
 
     const GLenum primitive_mode = MaxwellToGL::PrimitiveTopology(maxwell3d->regs.draw.topology);
     BeginTransformFeedback(pipeline, primitive_mode);
 
     const GLuint base_instance = static_cast<GLuint>(maxwell3d->regs.global_base_instance_index);
-    const GLsizei num_instances =
-        static_cast<GLsizei>(is_instanced ? maxwell3d->mme_draw.instance_count : 1);
+    const GLsizei num_instances = static_cast<GLsizei>(instance_count);
     if (is_indexed) {
         const GLint base_vertex = static_cast<GLint>(maxwell3d->regs.global_base_vertex_index);
         const GLsizei num_vertices = static_cast<GLsizei>(maxwell3d->regs.index_buffer.count);
@@ -1129,6 +1130,16 @@ void RasterizerOpenGL::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
+void RasterizerOpenGL::BindInlineIndexBuffer() {
+    if (maxwell3d->inline_index_draw_indexes.empty()) {
+        return;
+    }
+    const auto data_count = static_cast<u32>(maxwell3d->inline_index_draw_indexes.size());
+    auto buffer = Buffer(buffer_cache_runtime, *this, 0, data_count);
+    buffer.ImmediateUpload(0, maxwell3d->inline_index_draw_indexes);
+    buffer_cache_runtime.BindIndexBuffer(buffer, 0, data_count);
+}
+
 AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_) : buffer_cache{buffer_cache_} {}
 
 bool AccelerateDMA::BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -68,7 +68,7 @@ public:
                               StateTracker& state_tracker_);
     ~RasterizerOpenGL() override;
 
-    void Draw(bool is_indexed, bool is_instanced) override;
+    void Draw(bool is_indexed, u32 instance_count) override;
     void Clear() override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -199,6 +199,8 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
+    void BindInlineIndexBuffer();
+
     Tegra::GPU& gpu;
 
     const Device& device;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -63,6 +63,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(const GraphicsPipelineKey& key,
     Shader::RuntimeInfo info;
     if (previous_program) {
         info.previous_stage_stores = previous_program->info.stores;
+        info.previous_stage_legacy_stores_mapping = previous_program->info.legacy_stores_mapping;
     } else {
         // Mark all stores as available for vertex shaders
         info.previous_stage_stores.mask.set();

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -134,6 +134,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
     Shader::RuntimeInfo info;
     if (previous_program) {
         info.previous_stage_stores = previous_program->info.stores;
+        info.previous_stage_legacy_stores_mapping = previous_program->info.legacy_stores_mapping;
         if (previous_program->is_geometry_passthrough) {
             info.previous_stage_stores.mask |= previous_program->info.passthrough.mask;
         }

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -127,11 +127,10 @@ VkRect2D GetScissorState(const Maxwell& regs, size_t index, u32 up_scale = 1, u3
     return scissor;
 }
 
-DrawParams MakeDrawParams(const Maxwell& regs, u32 num_instances, bool is_instanced,
-                          bool is_indexed) {
+DrawParams MakeDrawParams(const Maxwell& regs, u32 num_instances, bool is_indexed) {
     DrawParams params{
         .base_instance = regs.global_base_instance_index,
-        .num_instances = is_instanced ? num_instances : 1,
+        .num_instances = num_instances,
         .base_vertex = is_indexed ? regs.global_base_vertex_index : regs.vertex_buffer.first,
         .num_vertices = is_indexed ? regs.index_buffer.count : regs.vertex_buffer.count,
         .first_index = is_indexed ? regs.index_buffer.first : 0,
@@ -177,7 +176,7 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
 
 RasterizerVulkan::~RasterizerVulkan() = default;
 
-void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
+void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
     MICROPROFILE_SCOPE(Vulkan_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
@@ -194,13 +193,15 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
     pipeline->SetEngine(maxwell3d, gpu_memory);
     pipeline->Configure(is_indexed);
 
+    BindInlineIndexBuffer();
+
     BeginTransformFeedback();
 
     UpdateDynamicStates();
 
     const auto& regs{maxwell3d->regs};
-    const u32 num_instances{maxwell3d->mme_draw.instance_count};
-    const DrawParams draw_params{MakeDrawParams(regs, num_instances, is_instanced, is_indexed)};
+    const u32 num_instances{instance_count};
+    const DrawParams draw_params{MakeDrawParams(regs, num_instances, is_indexed)};
     scheduler.Record([draw_params](vk::CommandBuffer cmdbuf) {
         if (draw_params.is_indexed) {
             cmdbuf.DrawIndexed(draw_params.num_vertices, draw_params.num_instances,
@@ -1009,4 +1010,17 @@ void RasterizerVulkan::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
+void RasterizerVulkan::BindInlineIndexBuffer() {
+    if (maxwell3d->inline_index_draw_indexes.empty()) {
+        return;
+    }
+    const auto data_count = static_cast<u32>(maxwell3d->inline_index_draw_indexes.size());
+    auto buffer = buffer_cache_runtime.UploadStagingBuffer(data_count);
+    std::memcpy(buffer.mapped_span.data(), maxwell3d->inline_index_draw_indexes.data(), data_count);
+    buffer_cache_runtime.BindIndexBuffer(
+        maxwell3d->regs.draw.topology, maxwell3d->regs.index_buffer.format,
+        maxwell3d->regs.index_buffer.first, maxwell3d->regs.index_buffer.count, buffer.buffer,
+        static_cast<u32>(buffer.offset), data_count);
+}
+
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -64,7 +64,7 @@ public:
                               StateTracker& state_tracker_, Scheduler& scheduler_);
     ~RasterizerVulkan() override;
 
-    void Draw(bool is_indexed, bool is_instanced) override;
+    void Draw(bool is_indexed, u32 instance_count) override;
     void Clear() override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -141,6 +141,8 @@ private:
 
     void UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs);
 
+    void BindInlineIndexBuffer();
+
     Tegra::GPU& gpu;
 
     ScreenInfo& screen_info;

src/video_core/texture_cache/texture_cache.h

@@ -442,7 +442,7 @@ void TextureCache<P>::WriteMemory(VAddr cpu_addr, size_t size) {
 template <class P>
 void TextureCache<P>::DownloadMemory(VAddr cpu_addr, size_t size) {
     std::vector<ImageId> images;
-    ForEachImageInRegion(cpu_addr, size, [this, &images](ImageId image_id, ImageBase& image) {
+    ForEachImageInRegion(cpu_addr, size, [&images](ImageId image_id, ImageBase& image) {
         if (!image.IsSafeDownload()) {
             return;
         }
@@ -1502,9 +1502,9 @@ void TextureCache<P>::UnregisterImage(ImageId image_id) {
     image.flags &= ~ImageFlagBits::BadOverlap;
     lru_cache.Free(image.lru_index);
     const auto& clear_page_table =
-        [this, image_id](u64 page,
-                         std::unordered_map<u64, std::vector<ImageId>, Common::IdentityHash<u64>>&
-                             selected_page_table) {
+        [image_id](u64 page,
+                   std::unordered_map<u64, std::vector<ImageId>, Common::IdentityHash<u64>>&
+                       selected_page_table) {
             const auto page_it = selected_page_table.find(page);
             if (page_it == selected_page_table.end()) {
                 ASSERT_MSG(false, "Unregistering unregistered page=0x{:x}", page << YUZU_PAGEBITS);

src/video_core/textures/astc.cpp

@@ -1661,8 +1661,8 @@ void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height,
     for (u32 z = 0; z < depth; ++z) {
         const u32 depth_offset = z * height * width * 4;
         for (u32 y_index = 0; y_index < rows; ++y_index) {
-            auto decompress_stride = [data, width, height, depth, block_width, block_height, output,
-                                      rows, cols, z, depth_offset, y_index] {
+            auto decompress_stride = [data, width, height, block_width, block_height, output, rows,
+                                      cols, z, depth_offset, y_index] {
                 const u32 y = y_index * block_height;
                 for (u32 x_index = 0; x_index < cols; ++x_index) {
                     const u32 block_index = (z * rows * cols) + (y_index * cols) + x_index;

src/video_core/textures/decoders.cpp

@@ -21,7 +21,7 @@ constexpr u32 pdep(u32 value) {
     u32 m = mask;
     for (u32 bit = 1; m; bit += bit) {
         if (value & bit)
-            result |= m & -m;
+            result |= m & (~m + 1);
         m &= m - 1;
     }
     return result;

src/yuzu/multiplayer/state.cpp

@@ -268,7 +268,7 @@ bool MultiplayerState::OnCloseRoom() {
             return true;
         }
         // Save ban list
-        UISettings::values.multiplayer_ban_list = std::move(room->GetBanList());
+        UISettings::values.multiplayer_ban_list = room->GetBanList();
 
         room->Destroy();
         announce_multiplayer_session->Stop();

src/yuzu/startup_checks.cpp

@@ -49,7 +49,7 @@ bool CheckEnvVars(bool* is_child) {
         *is_child = true;
         return false;
     } else if (!SetEnvironmentVariableA(IS_CHILD_ENV_VAR, ENV_VAR_ENABLED_TEXT)) {
-        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %d\n",
+        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %lu\n",
                      IS_CHILD_ENV_VAR, GetLastError());
         return true;
     }
@@ -62,7 +62,7 @@ bool StartupChecks(const char* arg0, bool* has_broken_vulkan, bool perform_vulka
     // Set the startup variable for child processes
     const bool env_var_set = SetEnvironmentVariableA(STARTUP_CHECK_ENV_VAR, ENV_VAR_ENABLED_TEXT);
     if (!env_var_set) {
-        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %d\n",
+        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %lu\n",
                      STARTUP_CHECK_ENV_VAR, GetLastError());
         return false;
     }
@@ -81,22 +81,22 @@ bool StartupChecks(const char* arg0, bool* has_broken_vulkan, bool perform_vulka
         DWORD exit_code = STILL_ACTIVE;
         const int err = GetExitCodeProcess(process_info.hProcess, &exit_code);
         if (err == 0) {
-            std::fprintf(stderr, "GetExitCodeProcess failed with error %d\n", GetLastError());
+            std::fprintf(stderr, "GetExitCodeProcess failed with error %lu\n", GetLastError());
         }
 
         // Vulkan is broken if the child crashed (return value is not zero)
         *has_broken_vulkan = (exit_code != 0);
 
         if (CloseHandle(process_info.hProcess) == 0) {
-            std::fprintf(stderr, "CloseHandle failed with error %d\n", GetLastError());
+            std::fprintf(stderr, "CloseHandle failed with error %lu\n", GetLastError());
         }
         if (CloseHandle(process_info.hThread) == 0) {
-            std::fprintf(stderr, "CloseHandle failed with error %d\n", GetLastError());
+            std::fprintf(stderr, "CloseHandle failed with error %lu\n", GetLastError());
         }
     }
 
     if (!SetEnvironmentVariableA(STARTUP_CHECK_ENV_VAR, nullptr)) {
-        std::fprintf(stderr, "SetEnvironmentVariableA failed to clear %s with error %d\n",
+        std::fprintf(stderr, "SetEnvironmentVariableA failed to clear %s with error %lu\n",
                      STARTUP_CHECK_ENV_VAR, GetLastError());
     }
 
@@ -135,7 +135,8 @@ bool SpawnChild(const char* arg0, PROCESS_INFORMATION* pi, int flags) {
     startup_info.cb = sizeof(startup_info);
 
     char p_name[255];
-    std::strncpy(p_name, arg0, 255);
+    std::strncpy(p_name, arg0, 254);
+    p_name[254] = '\0';
 
     const bool process_created = CreateProcessA(nullptr,       // lpApplicationName
                                                 p_name,        // lpCommandLine
@@ -149,7 +150,7 @@ bool SpawnChild(const char* arg0, PROCESS_INFORMATION* pi, int flags) {
                                                 pi             // lpProcessInformation
     );
     if (!process_created) {
-        std::fprintf(stderr, "CreateProcessA failed with error %d\n", GetLastError());
+        std::fprintf(stderr, "CreateProcessA failed with error %lu\n", GetLastError());
         return false;
     }