video_core: Enable ImageGather rounding fix on AMD open source drivers

Should be more efficient and better readable

src/audio_core/device/device_session.cpp

@@ -121,8 +121,7 @@ u64 DeviceSession::GetPlayedSampleCount() const {
 }
 
 std::optional<std::chrono::nanoseconds> DeviceSession::ThreadFunc() {
-    // Add 5ms of samples at a 48K sample rate.
-    played_sample_count += 48'000 * INCREMENT_TIME / 1s;
+    played_sample_count = stream->GetExpectedPlayedSampleCount();
     if (type == Sink::StreamType::Out) {
         system.AudioCore().GetAudioManager().SetEvent(Event::Type::AudioOutManager, true);
     } else {

src/audio_core/renderer/adsp/audio_renderer.cpp

@@ -189,6 +189,8 @@ void AudioRenderer::ThreadFunc() {
                     max_time = std::min(command_buffer.time_limit, max_time);
                     command_list_processor.SetProcessTimeMax(max_time);
 
+                    streams[index]->WaitFreeSpace();
+
                     // Process the command list
                     {
                         MICROPROFILE_SCOPE(Audio_Renderer);

src/audio_core/renderer/adsp/audio_renderer.h

@@ -10,6 +10,7 @@
 #include "audio_core/renderer/adsp/command_buffer.h"
 #include "audio_core/renderer/adsp/command_list_processor.h"
 #include "common/common_types.h"
+#include "common/polyfill_thread.h"
 #include "common/reader_writer_queue.h"
 #include "common/thread.h"
 

src/audio_core/renderer/system_manager.cpp

@@ -15,14 +15,9 @@ MICROPROFILE_DEFINE(Audio_RenderSystemManager, "Audio", "Render System Manager",
                     MP_RGB(60, 19, 97));
 
 namespace AudioCore::AudioRenderer {
-constexpr std::chrono::nanoseconds RENDER_TIME{5'000'000UL};
 
 SystemManager::SystemManager(Core::System& core_)
-    : core{core_}, adsp{core.AudioCore().GetADSP()}, mailbox{adsp.GetRenderMailbox()},
-      thread_event{Core::Timing::CreateEvent(
-          "AudioRendererSystemManager", [this](std::uintptr_t, s64 time, std::chrono::nanoseconds) {
-              return ThreadFunc2(time);
-          })} {}
+    : core{core_}, adsp{core.AudioCore().GetADSP()}, mailbox{adsp.GetRenderMailbox()} {}
 
 SystemManager::~SystemManager() {
     Stop();
@@ -33,8 +28,6 @@ bool SystemManager::InitializeUnsafe() {
         if (adsp.Start()) {
             active = true;
             thread = std::jthread([this](std::stop_token stop_token) { ThreadFunc(); });
-            core.CoreTiming().ScheduleLoopingEvent(std::chrono::nanoseconds(0), RENDER_TIME,
-                                                   thread_event);
         }
     }
 
@@ -45,7 +38,6 @@ void SystemManager::Stop() {
     if (!active) {
         return;
     }
-    core.CoreTiming().UnscheduleEvent(thread_event, {});
     active = false;
     update.store(true);
     update.notify_all();
@@ -111,16 +103,7 @@ void SystemManager::ThreadFunc() {
 
         adsp.Signal();
         adsp.Wait();
-
-        update.wait(false);
-        update.store(false);
     }
 }
 
-std::optional<std::chrono::nanoseconds> SystemManager::ThreadFunc2(s64 time) {
-    update.store(true);
-    update.notify_all();
-    return std::nullopt;
-}
-
 } // namespace AudioCore::AudioRenderer

src/audio_core/renderer/system_manager.h

@@ -68,11 +68,6 @@ private:
      */
     void ThreadFunc();
 
-    /**
-     * Signalling core timing thread to run ThreadFunc.
-     */
-    std::optional<std::chrono::nanoseconds> ThreadFunc2(s64 time);
-
     enum class StreamState {
         Filling,
         Steady,
@@ -95,8 +90,6 @@ private:
     ADSP::ADSP& adsp;
     /// AudioRenderer mailbox for communication
     ADSP::AudioRenderer_Mailbox* mailbox{};
-    /// Core timing event to signal main thread
-    std::shared_ptr<Core::Timing::EventType> thread_event;
     /// Atomic for main thread to wait on
     std::atomic<bool> update{};
 };

src/audio_core/sink/cubeb_sink.cpp

@@ -101,8 +101,6 @@ public:
     ~CubebSinkStream() override {
         LOG_DEBUG(Service_Audio, "Destructing cubeb stream {}", name);
 
-        Unstall();
-
         if (!ctx) {
             return;
         }
@@ -143,8 +141,6 @@ public:
      * Stop the sink stream.
      */
     void Stop() override {
-        Unstall();
-
         if (!ctx || paused) {
             return;
         }

src/audio_core/sink/sdl2_sink.cpp

@@ -88,7 +88,6 @@ public:
      * Finalize the sink stream.
      */
     void Finalize() override {
-        Unstall();
         if (device == 0) {
             return;
         }
@@ -116,7 +115,6 @@ public:
      * Stop the sink stream.
      */
     void Stop() override {
-        Unstall();
         if (device == 0 || paused) {
             return;
         }

src/audio_core/sink/sink_stream.cpp

@@ -14,6 +14,8 @@
 #include "common/fixed_point.h"
 #include "common/settings.h"
 #include "core/core.h"
+#include "core/core_timing.h"
+#include "core/core_timing_util.h"
 
 namespace AudioCore::Sink {
 
@@ -149,10 +151,6 @@ void SinkStream::ProcessAudioIn(std::span<const s16> input_buffer, std::size_t n
         return;
     }
 
-    if (queued_buffers > max_queue_size) {
-        Stall();
-    }
-
     while (frames_written < num_frames) {
         // If the playing buffer has been consumed or has no frames, we need a new one
         if (playing_buffer.consumed || playing_buffer.frames == 0) {
@@ -187,10 +185,6 @@ void SinkStream::ProcessAudioIn(std::span<const s16> input_buffer, std::size_t n
     }
 
     std::memcpy(&last_frame[0], &input_buffer[(frames_written - 1) * frame_size], frame_size_bytes);
-
-    if (queued_buffers <= max_queue_size) {
-        Unstall();
-    }
 }
 
 void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::size_t num_frames) {
@@ -198,10 +192,15 @@ void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::siz
     const std::size_t frame_size = num_channels;
     const std::size_t frame_size_bytes = frame_size * sizeof(s16);
     size_t frames_written{0};
+    size_t actual_frames_written{0};
 
     // If we're paused or going to shut down, we don't want to consume buffers as coretiming is
     // paused and we'll desync, so just play silence.
     if (system.IsPaused() || system.IsShuttingDown()) {
+        if (system.IsShuttingDown()) {
+            release_cv.notify_one();
+        }
+
         static constexpr std::array<s16, 6> silence{};
         for (size_t i = frames_written; i < num_frames; i++) {
             std::memcpy(&output_buffer[i * frame_size], &silence[0], frame_size_bytes);
@@ -209,20 +208,6 @@ void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::siz
         return;
     }
 
-    // Due to many frames being queued up with nvdec (5 frames or so?), a lot of buffers also get
-    // queued up (30+) but not all at once, which causes constant stalling here, so just let the
-    // video play out without attempting to stall.
-    // Can hopefully remove this later with a more complete NVDEC implementation.
-    const auto nvdec_active{system.AudioCore().IsNVDECActive()};
-
-    // Core timing cannot be paused in single-core mode, so Stall ends up being called over and over
-    // and never recovers to a normal state, so just skip attempting to sync things on single-core.
-    if (system.IsMulticore() && !nvdec_active && queued_buffers > max_queue_size) {
-        Stall();
-    } else if (system.IsMulticore() && queued_buffers <= max_queue_size) {
-        Unstall();
-    }
-
     while (frames_written < num_frames) {
         // If the playing buffer has been consumed or has no frames, we need a new one
         if (playing_buffer.consumed || playing_buffer.frames == 0) {
@@ -237,6 +222,10 @@ void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::siz
             }
             // Successfully dequeued a new buffer.
             queued_buffers--;
+
+            { std::unique_lock lk{release_mutex}; }
+
+            release_cv.notify_one();
         }
 
         // Get the minimum frames available between the currently playing buffer, and the
@@ -248,6 +237,7 @@ void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::siz
                            frames_available * frame_size);
 
         frames_written += frames_available;
+        actual_frames_written += frames_available;
         playing_buffer.frames_played += frames_available;
 
         // If that's all the frames in the current buffer, add its samples and mark it as
@@ -260,26 +250,29 @@ void SinkStream::ProcessAudioOutAndRender(std::span<s16> output_buffer, std::siz
     std::memcpy(&last_frame[0], &output_buffer[(frames_written - 1) * frame_size],
                 frame_size_bytes);
 
-    if (system.IsMulticore() && queued_buffers <= max_queue_size) {
-        Unstall();
+    {
+        std::scoped_lock lk{sample_count_lock};
+        last_sample_count_update_time =
+            Core::Timing::CyclesToUs(system.CoreTiming().GetClockTicks());
+        min_played_sample_count = max_played_sample_count;
+        max_played_sample_count += actual_frames_written;
     }
 }
 
-void SinkStream::Stall() {
-    std::scoped_lock lk{stall_guard};
-    if (stalled_lock) {
-        return;
-    }
-    stalled_lock = system.StallApplication();
+u64 SinkStream::GetExpectedPlayedSampleCount() {
+    std::scoped_lock lk{sample_count_lock};
+    auto cur_time{Core::Timing::CyclesToUs(system.CoreTiming().GetClockTicks())};
+    auto time_delta{cur_time - last_sample_count_update_time};
+    auto exp_played_sample_count{min_played_sample_count +
+                                 (TargetSampleRate * time_delta) / std::chrono::seconds{1}};
+
+    return std::min<u64>(exp_played_sample_count, max_played_sample_count);
 }
 
-void SinkStream::Unstall() {
-    std::scoped_lock lk{stall_guard};
-    if (!stalled_lock) {
-        return;
-    }
-    system.UnstallApplication();
-    stalled_lock.unlock();
+void SinkStream::WaitFreeSpace() {
+    std::unique_lock lk{release_mutex};
+    release_cv.wait(
+        lk, [this]() { return queued_buffers < max_queue_size || system.IsShuttingDown(); });
 }
 
 } // namespace AudioCore::Sink

src/audio_core/sink/sink_stream.h

@@ -5,6 +5,7 @@
 
 #include <array>
 #include <atomic>
+#include <chrono>
 #include <memory>
 #include <mutex>
 #include <span>
@@ -14,6 +15,7 @@
 #include "common/common_types.h"
 #include "common/reader_writer_queue.h"
 #include "common/ring_buffer.h"
+#include "common/thread.h"
 
 namespace Core {
 class System;
@@ -53,9 +55,7 @@ struct SinkBuffer {
 class SinkStream {
 public:
     explicit SinkStream(Core::System& system_, StreamType type_) : system{system_}, type{type_} {}
-    virtual ~SinkStream() {
-        Unstall();
-    }
+    virtual ~SinkStream() {}
 
     /**
      * Finalize the sink stream.
@@ -201,14 +201,16 @@ public:
     void ProcessAudioOutAndRender(std::span<s16> output_buffer, std::size_t num_frames);
 
     /**
-     * Stall core processes if the audio thread falls too far behind.
+     * Get the total number of samples expected to have been played by this stream.
+     *
+     * @return The number of samples.
      */
-    void Stall();
+    u64 GetExpectedPlayedSampleCount();
 
     /**
-     * Unstall core processes.
+     * Waits for free space in the sample ring buffer
      */
-    void Unstall();
+    void WaitFreeSpace();
 
 protected:
     /// Core system
@@ -237,12 +239,21 @@ private:
     std::atomic<u32> queued_buffers{};
     /// The ring size for audio out buffers (usually 4, rarely 2 or 8)
     u32 max_queue_size{};
+    /// Locks access to sample count tracking info
+    std::mutex sample_count_lock;
+    /// Minimum number of total samples that have been played since the last callback
+    u64 min_played_sample_count{};
+    /// Maximum number of total samples that can be played since the last callback
+    u64 max_played_sample_count{};
+    /// The time the two above tracking variables were last written to
+    std::chrono::microseconds last_sample_count_update_time{};
     /// Set by the audio render/in/out system which uses this stream
     f32 system_volume{1.0f};
     /// Set via IAudioDevice service calls
     f32 device_volume{1.0f};
-    std::mutex stall_guard;
-    std::unique_lock<std::mutex> stalled_lock;
+    /// Signalled when ring buffer entries are consumed
+    std::condition_variable release_cv;
+    std::mutex release_mutex;
 };
 
 using SinkStreamPtr = std::unique_ptr<SinkStream>;

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -5,7 +5,6 @@
 #include <memory>
 #include <dynarmic/interface/A32/a32.h>
 #include <dynarmic/interface/A32/config.h>
-#include <dynarmic/interface/A32/context.h>
 #include "common/assert.h"
 #include "common/literals.h"
 #include "common/logging/log.h"
@@ -410,21 +409,19 @@ void ARM_Dynarmic_32::SetTPIDR_EL0(u64 value) {
 }
 
 void ARM_Dynarmic_32::SaveContext(ThreadContext32& ctx) {
-    Dynarmic::A32::Context context;
-    jit.load()->SaveContext(context);
-    ctx.cpu_registers = context.Regs();
-    ctx.extension_registers = context.ExtRegs();
-    ctx.cpsr = context.Cpsr();
-    ctx.fpscr = context.Fpscr();
+    Dynarmic::A32::Jit* j = jit.load();
+    ctx.cpu_registers = j->Regs();
+    ctx.extension_registers = j->ExtRegs();
+    ctx.cpsr = j->Cpsr();
+    ctx.fpscr = j->Fpscr();
 }
 
 void ARM_Dynarmic_32::LoadContext(const ThreadContext32& ctx) {
-    Dynarmic::A32::Context context;
-    context.Regs() = ctx.cpu_registers;
-    context.ExtRegs() = ctx.extension_registers;
-    context.SetCpsr(ctx.cpsr);
-    context.SetFpscr(ctx.fpscr);
-    jit.load()->LoadContext(context);
+    Dynarmic::A32::Jit* j = jit.load();
+    j->Regs() = ctx.cpu_registers;
+    j->ExtRegs() = ctx.extension_registers;
+    j->SetCpsr(ctx.cpsr);
+    j->SetFpscr(ctx.fpscr);
 }
 
 void ARM_Dynarmic_32::SignalInterrupt() {

src/core/frontend/applets/controller.cpp

@@ -71,7 +71,7 @@ void DefaultControllerApplet::ReconfigureControllers(ReconfigureCallback callbac
         }
     }
 
-    callback();
+    callback(true);
 }
 
 } // namespace Core::Frontend

src/core/frontend/applets/controller.h

@@ -37,7 +37,7 @@ struct ControllerParameters {
 
 class ControllerApplet : public Applet {
 public:
-    using ReconfigureCallback = std::function<void()>;
+    using ReconfigureCallback = std::function<void(bool)>;
 
     virtual ~ControllerApplet();
 

src/core/frontend/applets/profile_select.cpp

@@ -11,7 +11,8 @@ ProfileSelectApplet::~ProfileSelectApplet() = default;
 
 void DefaultProfileSelectApplet::Close() const {}
 
-void DefaultProfileSelectApplet::SelectProfile(SelectProfileCallback callback) const {
+void DefaultProfileSelectApplet::SelectProfile(SelectProfileCallback callback,
+                                               const ProfileSelectParameters& parameters) const {
     Service::Account::ProfileManager manager;
     callback(manager.GetUser(Settings::values.current_user.GetValue()).value_or(Common::UUID{}));
     LOG_INFO(Service_ACC, "called, selecting current user instead of prompting...");

src/core/frontend/applets/profile_select.h

@@ -5,25 +5,35 @@
 
 #include <functional>
 #include <optional>
-#include "common/uuid.h"
 
+#include "common/uuid.h"
 #include "core/frontend/applets/applet.h"
+#include "core/hle/service/am/applets/applet_profile_select.h"
 
 namespace Core::Frontend {
 
+struct ProfileSelectParameters {
+    Service::AM::Applets::UiMode mode;
+    std::array<Common::UUID, 8> invalid_uid_list;
+    Service::AM::Applets::UiSettingsDisplayOptions display_options;
+    Service::AM::Applets::UserSelectionPurpose purpose;
+};
+
 class ProfileSelectApplet : public Applet {
 public:
     using SelectProfileCallback = std::function<void(std::optional<Common::UUID>)>;
 
     virtual ~ProfileSelectApplet();
 
-    virtual void SelectProfile(SelectProfileCallback callback) const = 0;
+    virtual void SelectProfile(SelectProfileCallback callback,
+                               const ProfileSelectParameters& parameters) const = 0;
 };
 
 class DefaultProfileSelectApplet final : public ProfileSelectApplet {
 public:
     void Close() const override;
-    void SelectProfile(SelectProfileCallback callback) const override;
+    void SelectProfile(SelectProfileCallback callback,
+                       const ProfileSelectParameters& parameters) const override;
 };
 
 } // namespace Core::Frontend

src/core/hle/kernel/k_address_arbiter.cpp

@@ -35,24 +35,30 @@ bool DecrementIfLessThan(Core::System& system, s32* out, KProcessAddress address
 
     // TODO(bunnei): We should call CanAccessAtomic(..) here.
 
-    // Load the value from the address.
-    const s32 current_value =
-        static_cast<s32>(monitor.ExclusiveRead32(current_core, GetInteger(address)));
+    s32 current_value{};
 
-    // Compare it to the desired one.
-    if (current_value < value) {
-        // If less than, we want to try to decrement.
-        const s32 decrement_value = current_value - 1;
+    while (true) {
+        // Load the value from the address.
+        current_value =
+            static_cast<s32>(monitor.ExclusiveRead32(current_core, GetInteger(address)));
+
+        // Compare it to the desired one.
+        if (current_value < value) {
+            // If less than, we want to try to decrement.
+            const s32 decrement_value = current_value - 1;
+
+            // Decrement and try to store.
+            if (monitor.ExclusiveWrite32(current_core, GetInteger(address),
+                                         static_cast<u32>(decrement_value))) {
+                break;
+            }
 
-        // Decrement and try to store.
-        if (!monitor.ExclusiveWrite32(current_core, GetInteger(address),
-                                      static_cast<u32>(decrement_value))) {
             // If we failed to store, try again.
-            DecrementIfLessThan(system, out, address, value);
+        } else {
+            // Otherwise, clear our exclusive hold and finish
+            monitor.ClearExclusive(current_core);
+            break;
         }
-    } else {
-        // Otherwise, clear our exclusive hold and finish
-        monitor.ClearExclusive(current_core);
     }
 
     // We're done.
@@ -70,23 +76,29 @@ bool UpdateIfEqual(Core::System& system, s32* out, KProcessAddress address, s32
 
     // TODO(bunnei): We should call CanAccessAtomic(..) here.
 
+    s32 current_value{};
+
     // Load the value from the address.
-    const s32 current_value =
-        static_cast<s32>(monitor.ExclusiveRead32(current_core, GetInteger(address)));
+    while (true) {
+        current_value =
+            static_cast<s32>(monitor.ExclusiveRead32(current_core, GetInteger(address)));
 
-    // Compare it to the desired one.
-    if (current_value == value) {
-        // If equal, we want to try to write the new value.
+        // Compare it to the desired one.
+        if (current_value == value) {
+            // If equal, we want to try to write the new value.
+
+            // Try to store.
+            if (monitor.ExclusiveWrite32(current_core, GetInteger(address),
+                                         static_cast<u32>(new_value))) {
+                break;
+            }
 
-        // Try to store.
-        if (!monitor.ExclusiveWrite32(current_core, GetInteger(address),
-                                      static_cast<u32>(new_value))) {
             // If we failed to store, try again.
-            UpdateIfEqual(system, out, address, value, new_value);
+        } else {
+            // Otherwise, clear our exclusive hold and finish.
+            monitor.ClearExclusive(current_core);
+            break;
         }
-    } else {
-        // Otherwise, clear our exclusive hold and finish.
-        monitor.ClearExclusive(current_core);
     }
 
     // We're done.

src/core/hle/kernel/k_condition_variable.cpp

@@ -33,21 +33,26 @@ bool UpdateLockAtomic(Core::System& system, u32* out, KProcessAddress address, u
     auto& monitor = system.Monitor();
     const auto current_core = system.Kernel().CurrentPhysicalCoreIndex();
 
-    // Load the value from the address.
-    const auto expected = monitor.ExclusiveRead32(current_core, GetInteger(address));
+    u32 expected{};
 
-    // Orr in the new mask.
-    u32 value = expected | new_orr_mask;
+    while (true) {
+        // Load the value from the address.
+        expected = monitor.ExclusiveRead32(current_core, GetInteger(address));
 
-    // If the value is zero, use the if_zero value, otherwise use the newly orr'd value.
-    if (!expected) {
-        value = if_zero;
-    }
+        // Orr in the new mask.
+        u32 value = expected | new_orr_mask;
+
+        // If the value is zero, use the if_zero value, otherwise use the newly orr'd value.
+        if (!expected) {
+            value = if_zero;
+        }
+
+        // Try to store.
+        if (monitor.ExclusiveWrite32(current_core, GetInteger(address), value)) {
+            break;
+        }
 
-    // Try to store.
-    if (!monitor.ExclusiveWrite32(current_core, GetInteger(address), value)) {
         // If we failed to store, try again.
-        return UpdateLockAtomic(system, out, address, if_zero, new_orr_mask);
     }
 
     // We're done.

src/core/hle/service/am/applets/applet_controller.cpp

@@ -224,7 +224,8 @@ void Controller::Execute() {
                   parameters.allow_dual_joycons, parameters.allow_left_joycon,
                   parameters.allow_right_joycon);
 
-        frontend.ReconfigureControllers([this] { ConfigurationComplete(); }, parameters);
+        frontend.ReconfigureControllers(
+            [this](bool is_success) { ConfigurationComplete(is_success); }, parameters);
         break;
     }
     case ControllerSupportMode::ShowControllerStrapGuide:
@@ -232,16 +233,16 @@ void Controller::Execute() {
     case ControllerSupportMode::ShowControllerKeyRemappingForSystem:
         UNIMPLEMENTED_MSG("ControllerSupportMode={} is not implemented",
                           controller_private_arg.mode);
-        ConfigurationComplete();
+        ConfigurationComplete(true);
         break;
     default: {
-        ConfigurationComplete();
+        ConfigurationComplete(true);
         break;
     }
     }
 }
 
-void Controller::ConfigurationComplete() {
+void Controller::ConfigurationComplete(bool is_success) {
     ControllerSupportResultInfo result_info{};
 
     // If enable_single_mode is enabled, player_count is 1 regardless of any other parameters.
@@ -250,7 +251,8 @@ void Controller::ConfigurationComplete() {
 
     result_info.selected_id = static_cast<u32>(system.HIDCore().GetFirstNpadId());
 
-    result_info.result = 0;
+    result_info.result =
+        is_success ? ControllerSupportResult::Success : ControllerSupportResult::Cancel;
 
     LOG_DEBUG(Service_HID, "Result Info: player_count={}, selected_id={}, result={}",
               result_info.player_count, result_info.selected_id, result_info.result);

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

@@ -48,6 +48,11 @@ enum class ControllerSupportCaller : u8 {
     MaxControllerSupportCaller,
 };
 
+enum class ControllerSupportResult : u32 {
+    Success = 0,
+    Cancel = 2,
+};
+
 struct ControllerSupportArgPrivate {
     u32 arg_private_size{};
     u32 arg_size{};
@@ -112,7 +117,7 @@ struct ControllerSupportResultInfo {
     s8 player_count{};
     INSERT_PADDING_BYTES(3);
     u32 selected_id{};
-    u32 result{};
+    ControllerSupportResult result{};
 };
 static_assert(sizeof(ControllerSupportResultInfo) == 0xC,
               "ControllerSupportResultInfo has incorrect size.");
@@ -131,7 +136,7 @@ public:
     void Execute() override;
     Result RequestExit() override;
 
-    void ConfigurationComplete();
+    void ConfigurationComplete(bool is_success);
 
 private:
     const Core::Frontend::ControllerApplet& frontend;

src/core/hle/service/am/applets/applet_profile_select.cpp

@@ -25,13 +25,29 @@ void ProfileSelect::Initialize() {
     final_data.clear();
 
     Applet::Initialize();
+    profile_select_version = ProfileSelectAppletVersion{common_args.library_version};
 
     const auto user_config_storage = broker.PopNormalDataToApplet();
     ASSERT(user_config_storage != nullptr);
     const auto& user_config = user_config_storage->GetData();
 
-    ASSERT(user_config.size() >= sizeof(UserSelectionConfig));
-    std::memcpy(&config, user_config.data(), sizeof(UserSelectionConfig));
+    LOG_INFO(Service_AM, "Initializing Profile Select Applet with version={}",
+             profile_select_version);
+
+    switch (profile_select_version) {
+    case ProfileSelectAppletVersion::Version1:
+        ASSERT(user_config.size() == sizeof(UiSettingsV1));
+        std::memcpy(&config_old, user_config.data(), sizeof(UiSettingsV1));
+        break;
+    case ProfileSelectAppletVersion::Version2:
+    case ProfileSelectAppletVersion::Version3:
+        ASSERT(user_config.size() == sizeof(UiSettings));
+        std::memcpy(&config, user_config.data(), sizeof(UiSettings));
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unknown profile_select_version = {}", profile_select_version);
+        break;
+    }
 }
 
 bool ProfileSelect::TransactionComplete() const {
@@ -52,11 +68,37 @@ void ProfileSelect::Execute() {
         return;
     }
 
-    frontend.SelectProfile([this](std::optional<Common::UUID> uuid) { SelectionComplete(uuid); });
+    Core::Frontend::ProfileSelectParameters parameters{};
+
+    switch (profile_select_version) {
+    case ProfileSelectAppletVersion::Version1:
+        parameters = {
+            .mode = config_old.mode,
+            .invalid_uid_list = config_old.invalid_uid_list,
+            .display_options = config_old.display_options,
+            .purpose = UserSelectionPurpose::General,
+        };
+        break;
+    case ProfileSelectAppletVersion::Version2:
+    case ProfileSelectAppletVersion::Version3:
+        parameters = {
+            .mode = config.mode,
+            .invalid_uid_list = config.invalid_uid_list,
+            .display_options = config.display_options,
+            .purpose = config.purpose,
+        };
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unknown profile_select_version = {}", profile_select_version);
+        break;
+    }
+
+    frontend.SelectProfile([this](std::optional<Common::UUID> uuid) { SelectionComplete(uuid); },
+                           parameters);
 }
 
 void ProfileSelect::SelectionComplete(std::optional<Common::UUID> uuid) {
-    UserSelectionOutput output{};
+    UiReturnArg output{};
 
     if (uuid.has_value() && uuid->IsValid()) {
         output.result = 0;
@@ -67,7 +109,7 @@ void ProfileSelect::SelectionComplete(std::optional<Common::UUID> uuid) {
         output.uuid_selected = Common::InvalidUUID;
     }
 
-    final_data = std::vector<u8>(sizeof(UserSelectionOutput));
+    final_data = std::vector<u8>(sizeof(UiReturnArg));
     std::memcpy(final_data.data(), &output, final_data.size());
     broker.PushNormalDataFromApplet(std::make_shared<IStorage>(system, std::move(final_data)));
     broker.SignalStateChanged();

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

@@ -16,19 +16,100 @@ class System;
 
 namespace Service::AM::Applets {
 
-struct UserSelectionConfig {
-    // TODO(DarkLordZach): RE this structure
-    // It seems to be flags and the like that determine the UI of the applet on the switch... from
-    // my research this is safe to ignore for now.
-    INSERT_PADDING_BYTES(0xA0);
+enum class ProfileSelectAppletVersion : u32 {
+    Version1 = 0x1,     // 1.0.0+
+    Version2 = 0x10000, // 2.0.0+
+    Version3 = 0x20000, // 6.0.0+
 };
-static_assert(sizeof(UserSelectionConfig) == 0xA0, "UserSelectionConfig has incorrect size.");
 
-struct UserSelectionOutput {
+// This is nn::account::UiMode
+enum class UiMode {
+    UserSelector,
+    UserCreator,
+    EnsureNetworkServiceAccountAvailable,
+    UserIconEditor,
+    UserNicknameEditor,
+    UserCreatorForStarter,
+    NintendoAccountAuthorizationRequestContext,
+    IntroduceExternalNetworkServiceAccount,
+    IntroduceExternalNetworkServiceAccountForRegistration,
+    NintendoAccountNnidLinker,
+    LicenseRequirementsForNetworkService,
+    LicenseRequirementsForNetworkServiceWithUserContextImpl,
+    UserCreatorForImmediateNaLoginTest,
+    UserQualificationPromoter,
+};
+
+// This is nn::account::UserSelectionPurpose
+enum class UserSelectionPurpose {
+    General,
+    GameCardRegistration,
+    EShopLaunch,
+    EShopItemShow,
+    PicturePost,
+    NintendoAccountLinkage,
+    SettingsUpdate,
+    SaveDataDeletion,
+    UserMigration,
+    SaveDataTransfer,
+};
+
+// This is nn::account::NintendoAccountStartupDialogType
+enum class NintendoAccountStartupDialogType {
+    LoginAndCreate,
+    Login,
+    Create,
+};
+
+// This is nn::account::UserSelectionSettingsForSystemService
+struct UserSelectionSettingsForSystemService {
+    UserSelectionPurpose purpose;
+    bool enable_user_creation;
+    INSERT_PADDING_BYTES(0x3);
+};
+static_assert(sizeof(UserSelectionSettingsForSystemService) == 0x8,
+              "UserSelectionSettingsForSystemService has incorrect size.");
+
+struct UiSettingsDisplayOptions {
+    bool is_network_service_account_required;
+    bool is_skip_enabled;
+    bool is_system_or_launcher;
+    bool is_registration_permitted;
+    bool show_skip_button;
+    bool aditional_select;
+    bool show_user_selector;
+    bool is_unqualified_user_selectable;
+};
+static_assert(sizeof(UiSettingsDisplayOptions) == 0x8,
+              "UiSettingsDisplayOptions has incorrect size.");
+
+struct UiSettingsV1 {
+    UiMode mode;
+    INSERT_PADDING_BYTES(0x4);
+    std::array<Common::UUID, 8> invalid_uid_list;
+    u64 application_id;
+    UiSettingsDisplayOptions display_options;
+};
+static_assert(sizeof(UiSettingsV1) == 0x98, "UiSettings has incorrect size.");
+
+// This is nn::account::UiSettings
+struct UiSettings {
+    UiMode mode;
+    INSERT_PADDING_BYTES(0x4);
+    std::array<Common::UUID, 8> invalid_uid_list;
+    u64 application_id;
+    UiSettingsDisplayOptions display_options;
+    UserSelectionPurpose purpose;
+    INSERT_PADDING_BYTES(0x4);
+};
+static_assert(sizeof(UiSettings) == 0xA0, "UiSettings has incorrect size.");
+
+// This is nn::account::UiReturnArg
+struct UiReturnArg {
     u64 result;
     Common::UUID uuid_selected;
 };
-static_assert(sizeof(UserSelectionOutput) == 0x18, "UserSelectionOutput has incorrect size.");
+static_assert(sizeof(UiReturnArg) == 0x18, "UiReturnArg has incorrect size.");
 
 class ProfileSelect final : public Applet {
 public:
@@ -49,7 +130,10 @@ public:
 private:
     const Core::Frontend::ProfileSelectApplet& frontend;
 
-    UserSelectionConfig config;
+    UiSettings config;
+    UiSettingsV1 config_old;
+    ProfileSelectAppletVersion profile_select_version;
+
     bool complete = false;
     Result status = ResultSuccess;
     std::vector<u8> final_data;

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

@@ -70,7 +70,6 @@ Result Controller_NPad::VerifyValidSixAxisSensorHandle(
     const Core::HID::SixAxisSensorHandle& device_handle) {
     const auto npad_id = IsNpadIdValid(static_cast<Core::HID::NpadIdType>(device_handle.npad_id));
     const bool device_index = device_handle.device_index < Core::HID::DeviceIndex::MaxDeviceIndex;
-    const bool npad_type = device_handle.npad_type < Core::HID::NpadStyleIndex::MaxNpadType;
 
     if (!npad_id) {
         return InvalidNpadId;
@@ -78,10 +77,6 @@ Result Controller_NPad::VerifyValidSixAxisSensorHandle(
     if (!device_index) {
         return NpadDeviceIndexOutOfRange;
     }
-    // This doesn't get validated on nnsdk
-    if (!npad_type) {
-        return NpadInvalidHandle;
-    }
 
     return ResultSuccess;
 }
@@ -819,12 +814,12 @@ Controller_NPad::NpadCommunicationMode Controller_NPad::GetNpadCommunicationMode
     return communication_mode;
 }
 
-Result Controller_NPad::SetNpadMode(Core::HID::NpadIdType npad_id,
-                                    NpadJoyDeviceType npad_device_type,
-                                    NpadJoyAssignmentMode assignment_mode) {
+bool Controller_NPad::SetNpadMode(Core::HID::NpadIdType& new_npad_id, Core::HID::NpadIdType npad_id,
+                                  NpadJoyDeviceType npad_device_type,
+                                  NpadJoyAssignmentMode assignment_mode) {
     if (!IsNpadIdValid(npad_id)) {
         LOG_ERROR(Service_HID, "Invalid NpadIdType npad_id:{}", npad_id);
-        return InvalidNpadId;
+        return false;
     }
 
     auto& controller = GetControllerFromNpadIdType(npad_id);
@@ -833,7 +828,7 @@ Result Controller_NPad::SetNpadMode(Core::HID::NpadIdType npad_id,
     }
 
     if (!controller.device->IsConnected()) {
-        return ResultSuccess;
+        return false;
     }
 
     if (assignment_mode == NpadJoyAssignmentMode::Dual) {
@@ -842,52 +837,52 @@ Result Controller_NPad::SetNpadMode(Core::HID::NpadIdType npad_id,
             controller.is_dual_left_connected = true;
             controller.is_dual_right_connected = false;
             UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconDual, npad_id, true);
-            return ResultSuccess;
+            return false;
         }
         if (controller.device->GetNpadStyleIndex() == Core::HID::NpadStyleIndex::JoyconRight) {
             DisconnectNpad(npad_id);
             controller.is_dual_left_connected = false;
             controller.is_dual_right_connected = true;
             UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconDual, npad_id, true);
-            return ResultSuccess;
+            return false;
         }
-        return ResultSuccess;
+        return false;
     }
 
     // This is for NpadJoyAssignmentMode::Single
 
     // Only JoyconDual get affected by this function
     if (controller.device->GetNpadStyleIndex() != Core::HID::NpadStyleIndex::JoyconDual) {
-        return ResultSuccess;
+        return false;
     }
 
     if (controller.is_dual_left_connected && !controller.is_dual_right_connected) {
         DisconnectNpad(npad_id);
         UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconLeft, npad_id, true);
-        return ResultSuccess;
+        return false;
     }
     if (!controller.is_dual_left_connected && controller.is_dual_right_connected) {
         DisconnectNpad(npad_id);
         UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconRight, npad_id, true);
-        return ResultSuccess;
+        return false;
     }
 
     // We have two controllers connected to the same npad_id we need to split them
-    const auto npad_id_2 = hid_core.GetFirstDisconnectedNpadId();
-    auto& controller_2 = GetControllerFromNpadIdType(npad_id_2);
+    new_npad_id = hid_core.GetFirstDisconnectedNpadId();
+    auto& controller_2 = GetControllerFromNpadIdType(new_npad_id);
     DisconnectNpad(npad_id);
     if (npad_device_type == NpadJoyDeviceType::Left) {
         UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconLeft, npad_id, true);
         controller_2.is_dual_left_connected = false;
         controller_2.is_dual_right_connected = true;
-        UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconDual, npad_id_2, true);
+        UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconDual, new_npad_id, true);
     } else {
         UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconRight, npad_id, true);
         controller_2.is_dual_left_connected = true;
         controller_2.is_dual_right_connected = false;
-        UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconDual, npad_id_2, true);
+        UpdateControllerAt(Core::HID::NpadStyleIndex::JoyconDual, new_npad_id, true);
     }
-    return ResultSuccess;
+    return true;
 }
 
 bool Controller_NPad::VibrateControllerAtIndex(Core::HID::NpadIdType npad_id,
@@ -1131,6 +1126,7 @@ Result Controller_NPad::DisconnectNpad(Core::HID::NpadIdType npad_id) {
     WriteEmptyEntry(shared_memory);
     return ResultSuccess;
 }
+
 Result Controller_NPad::SetGyroscopeZeroDriftMode(
     const Core::HID::SixAxisSensorHandle& sixaxis_handle,
     Core::HID::GyroscopeZeroDriftMode drift_mode) {

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

@@ -102,8 +102,8 @@ public:
     void SetNpadCommunicationMode(NpadCommunicationMode communication_mode_);
     NpadCommunicationMode GetNpadCommunicationMode() const;
 
-    Result SetNpadMode(Core::HID::NpadIdType npad_id, NpadJoyDeviceType npad_device_type,
-                       NpadJoyAssignmentMode assignment_mode);
+    bool SetNpadMode(Core::HID::NpadIdType& new_npad_id, Core::HID::NpadIdType npad_id,
+                     NpadJoyDeviceType npad_device_type, NpadJoyAssignmentMode assignment_mode);
 
     bool VibrateControllerAtIndex(Core::HID::NpadIdType npad_id, std::size_t device_index,
                                   const Core::HID::VibrationValue& vibration_value);

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

@@ -302,7 +302,7 @@ Hid::Hid(Core::System& system_)
         {130, &Hid::SwapNpadAssignment, "SwapNpadAssignment"},
         {131, &Hid::IsUnintendedHomeButtonInputProtectionEnabled, "IsUnintendedHomeButtonInputProtectionEnabled"},
         {132, &Hid::EnableUnintendedHomeButtonInputProtection, "EnableUnintendedHomeButtonInputProtection"},
-        {133, nullptr, "SetNpadJoyAssignmentModeSingleWithDestination"},
+        {133, &Hid::SetNpadJoyAssignmentModeSingleWithDestination, "SetNpadJoyAssignmentModeSingleWithDestination"},
         {134, &Hid::SetNpadAnalogStickUseCenterClamp, "SetNpadAnalogStickUseCenterClamp"},
         {135, &Hid::SetNpadCaptureButtonAssignment, "SetNpadCaptureButtonAssignment"},
         {136, &Hid::ClearNpadCaptureButtonAssignment, "ClearNpadCaptureButtonAssignment"},
@@ -1180,8 +1180,10 @@ void Hid::SetNpadJoyAssignmentModeSingleByDefault(HLERequestContext& ctx) {
 
     const auto parameters{rp.PopRaw<Parameters>()};
 
+    Core::HID::NpadIdType new_npad_id{};
     auto& controller = GetAppletResource()->GetController<Controller_NPad>(HidController::NPad);
-    controller.SetNpadMode(parameters.npad_id, Controller_NPad::NpadJoyDeviceType::Left,
+    controller.SetNpadMode(new_npad_id, parameters.npad_id,
+                           Controller_NPad::NpadJoyDeviceType::Left,
                            Controller_NPad::NpadJoyAssignmentMode::Single);
 
     LOG_INFO(Service_HID, "called, npad_id={}, applet_resource_user_id={}", parameters.npad_id,
@@ -1203,8 +1205,9 @@ void Hid::SetNpadJoyAssignmentModeSingle(HLERequestContext& ctx) {
 
     const auto parameters{rp.PopRaw<Parameters>()};
 
+    Core::HID::NpadIdType new_npad_id{};
     auto& controller = GetAppletResource()->GetController<Controller_NPad>(HidController::NPad);
-    controller.SetNpadMode(parameters.npad_id, parameters.npad_joy_device_type,
+    controller.SetNpadMode(new_npad_id, parameters.npad_id, parameters.npad_joy_device_type,
                            Controller_NPad::NpadJoyAssignmentMode::Single);
 
     LOG_INFO(Service_HID, "called, npad_id={}, applet_resource_user_id={}, npad_joy_device_type={}",
@@ -1226,8 +1229,10 @@ void Hid::SetNpadJoyAssignmentModeDual(HLERequestContext& ctx) {
 
     const auto parameters{rp.PopRaw<Parameters>()};
 
+    Core::HID::NpadIdType new_npad_id{};
     auto& controller = GetAppletResource()->GetController<Controller_NPad>(HidController::NPad);
-    controller.SetNpadMode(parameters.npad_id, {}, Controller_NPad::NpadJoyAssignmentMode::Dual);
+    controller.SetNpadMode(new_npad_id, parameters.npad_id, {},
+                           Controller_NPad::NpadJoyAssignmentMode::Dual);
 
     LOG_INFO(Service_HID, "called, npad_id={}, applet_resource_user_id={}", parameters.npad_id,
              parameters.applet_resource_user_id);
@@ -1369,6 +1374,34 @@ void Hid::EnableUnintendedHomeButtonInputProtection(HLERequestContext& ctx) {
     rb.Push(result);
 }
 
+void Hid::SetNpadJoyAssignmentModeSingleWithDestination(HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    struct Parameters {
+        Core::HID::NpadIdType npad_id;
+        INSERT_PADDING_WORDS_NOINIT(1);
+        u64 applet_resource_user_id;
+        Controller_NPad::NpadJoyDeviceType npad_joy_device_type;
+    };
+    static_assert(sizeof(Parameters) == 0x18, "Parameters has incorrect size.");
+
+    const auto parameters{rp.PopRaw<Parameters>()};
+
+    Core::HID::NpadIdType new_npad_id{};
+    auto& controller = GetAppletResource()->GetController<Controller_NPad>(HidController::NPad);
+    const auto is_reassigned =
+        controller.SetNpadMode(new_npad_id, parameters.npad_id, parameters.npad_joy_device_type,
+                               Controller_NPad::NpadJoyAssignmentMode::Single);
+
+    LOG_INFO(Service_HID, "called, npad_id={}, applet_resource_user_id={}, npad_joy_device_type={}",
+             parameters.npad_id, parameters.applet_resource_user_id,
+             parameters.npad_joy_device_type);
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(ResultSuccess);
+    rb.Push(is_reassigned);
+    rb.PushEnum(new_npad_id);
+}
+
 void Hid::SetNpadAnalogStickUseCenterClamp(HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     struct Parameters {

src/core/hle/service/hid/hid.h

@@ -151,6 +151,7 @@ private:
     void SwapNpadAssignment(HLERequestContext& ctx);
     void IsUnintendedHomeButtonInputProtectionEnabled(HLERequestContext& ctx);
     void EnableUnintendedHomeButtonInputProtection(HLERequestContext& ctx);
+    void SetNpadJoyAssignmentModeSingleWithDestination(HLERequestContext& ctx);
     void SetNpadAnalogStickUseCenterClamp(HLERequestContext& ctx);
     void SetNpadCaptureButtonAssignment(HLERequestContext& ctx);
     void ClearNpadCaptureButtonAssignment(HLERequestContext& ctx);

src/shader_recompiler/backend/glsl/emit_glsl_image.cpp

@@ -143,6 +143,21 @@ IR::Inst* PrepareSparse(IR::Inst& inst) {
     }
     return sparse_inst;
 }
+
+std::string ImageGatherSubpixelOffset(const IR::TextureInstInfo& info, std::string_view texture,
+                                      std::string_view coords) {
+    switch (info.type) {
+    case TextureType::Color2D:
+    case TextureType::Color2DRect:
+        return fmt::format("{}+vec2(0.001953125)/vec2(textureSize({}, 0))", coords, texture);
+    case TextureType::ColorArray2D:
+    case TextureType::ColorCube:
+        return fmt::format("vec3({0}.xy+vec2(0.001953125)/vec2(textureSize({1}, 0)),{0}.z)", coords,
+                           texture);
+    default:
+        return std::string{coords};
+    }
+}
 } // Anonymous namespace
 
 void EmitImageSampleImplicitLod(EmitContext& ctx, IR::Inst& inst, const IR::Value& index,
@@ -340,6 +355,13 @@ void EmitImageGather(EmitContext& ctx, IR::Inst& inst, const IR::Value& index,
         LOG_WARNING(Shader_GLSL, "Device does not support sparse texture queries. STUBBING");
         ctx.AddU1("{}=true;", *sparse_inst);
     }
+    std::string coords_with_subpixel_offset;
+    if (ctx.profile.need_gather_subpixel_offset) {
+        // Apply a subpixel offset of 1/512 the texel size of the texture to ensure same rounding on
+        // AMD hardware as on Maxwell or other Nvidia architectures.
+        coords_with_subpixel_offset = ImageGatherSubpixelOffset(info, texture, coords);
+        coords = coords_with_subpixel_offset;
+    }
     if (!sparse_inst || !supports_sparse) {
         if (offset.IsEmpty()) {
             ctx.Add("{}=textureGather({},{},int({}));", texel, texture, coords,
@@ -387,6 +409,13 @@ void EmitImageGatherDref(EmitContext& ctx, IR::Inst& inst, const IR::Value& inde
         LOG_WARNING(Shader_GLSL, "Device does not support sparse texture queries. STUBBING");
         ctx.AddU1("{}=true;", *sparse_inst);
     }
+    std::string coords_with_subpixel_offset;
+    if (ctx.profile.need_gather_subpixel_offset) {
+        // Apply a subpixel offset of 1/512 the texel size of the texture to ensure same rounding on
+        // AMD hardware as on Maxwell or other Nvidia architectures.
+        coords_with_subpixel_offset = ImageGatherSubpixelOffset(info, texture, coords);
+        coords = coords_with_subpixel_offset;
+    }
     if (!sparse_inst || !supports_sparse) {
         if (offset.IsEmpty()) {
             ctx.Add("{}=textureGather({},{},{});", texel, texture, coords, dref);

src/shader_recompiler/backend/spirv/emit_spirv_image.cpp

@@ -261,6 +261,30 @@ Id BitTest(EmitContext& ctx, Id mask, Id bit) {
     const Id bit_value{ctx.OpBitwiseAnd(ctx.U32[1], shifted, ctx.Const(1u))};
     return ctx.OpINotEqual(ctx.U1, bit_value, ctx.u32_zero_value);
 }
+
+Id ImageGatherSubpixelOffset(EmitContext& ctx, const IR::TextureInstInfo& info, Id texture,
+                             Id coords) {
+    // Apply a subpixel offset of 1/512 the texel size of the texture to ensure same rounding on
+    // AMD hardware as on Maxwell or other Nvidia architectures.
+    const auto calculate_coords{[&](size_t dim) {
+        const Id nudge{ctx.Const(0x1p-9f)};
+        const Id image_size{ctx.OpImageQuerySizeLod(ctx.U32[dim], texture, ctx.u32_zero_value)};
+        Id offset{dim == 2 ? ctx.ConstantComposite(ctx.F32[dim], nudge, nudge)
+                           : ctx.ConstantComposite(ctx.F32[dim], nudge, nudge, ctx.f32_zero_value)};
+        offset = ctx.OpFDiv(ctx.F32[dim], offset, ctx.OpConvertUToF(ctx.F32[dim], image_size));
+        return ctx.OpFAdd(ctx.F32[dim], coords, offset);
+    }};
+    switch (info.type) {
+    case TextureType::Color2D:
+    case TextureType::Color2DRect:
+        return calculate_coords(2);
+    case TextureType::ColorArray2D:
+    case TextureType::ColorCube:
+        return calculate_coords(3);
+    default:
+        return coords;
+    }
+}
 } // Anonymous namespace
 
 Id EmitBindlessImageSampleImplicitLod(EmitContext&) {
@@ -423,6 +447,9 @@ Id EmitImageGather(EmitContext& ctx, IR::Inst* inst, const IR::Value& index, Id
                    const IR::Value& offset, const IR::Value& offset2) {
     const auto info{inst->Flags<IR::TextureInstInfo>()};
     const ImageOperands operands(ctx, offset, offset2);
+    if (ctx.profile.need_gather_subpixel_offset) {
+        coords = ImageGatherSubpixelOffset(ctx, info, TextureImage(ctx, info, index), coords);
+    }
     return Emit(&EmitContext::OpImageSparseGather, &EmitContext::OpImageGather, ctx, inst,
                 ctx.F32[4], Texture(ctx, info, index), coords, ctx.Const(info.gather_component),
                 operands.MaskOptional(), operands.Span());
@@ -432,6 +459,9 @@ Id EmitImageGatherDref(EmitContext& ctx, IR::Inst* inst, const IR::Value& index,
                        const IR::Value& offset, const IR::Value& offset2, Id dref) {
     const auto info{inst->Flags<IR::TextureInstInfo>()};
     const ImageOperands operands(ctx, offset, offset2);
+    if (ctx.profile.need_gather_subpixel_offset) {
+        coords = ImageGatherSubpixelOffset(ctx, info, TextureImage(ctx, info, index), coords);
+    }
     return Emit(&EmitContext::OpImageSparseDrefGather, &EmitContext::OpImageDrefGather, ctx, inst,
                 ctx.F32[4], Texture(ctx, info, index), coords, dref, operands.MaskOptional(),
                 operands.Span());

src/shader_recompiler/profile.h

@@ -52,6 +52,10 @@ struct Profile {
     bool need_declared_frag_colors{};
     /// Prevents fast math optimizations that may cause inaccuracies
     bool need_fastmath_off{};
+    /// Some GPU vendors use a different rounding precision when calculating texture pixel
+    /// coordinates with the 16.8 format in the ImageGather instruction than the Maxwell
+    /// architecture. Applying an offset does fix this mismatching rounding behaviour.
+    bool need_gather_subpixel_offset{};
 
     /// OpFClamp is broken and OpFMax + OpFMin should be used instead
     bool has_broken_spirv_clamp{};

src/video_core/renderer_opengl/gl_device.h

@@ -176,6 +176,10 @@ public:
         return vendor_name == "ATI Technologies Inc.";
     }
 
+    bool IsIntel() const {
+        return vendor_name == "Intel";
+    }
+
     bool CanReportMemoryUsage() const {
         return can_report_memory;
     }

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -218,6 +218,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .lower_left_origin_mode = true,
           .need_declared_frag_colors = true,
           .need_fastmath_off = device.NeedsFastmathOff(),
+          .need_gather_subpixel_offset = device.IsAmd() || device.IsIntel(),
 
           .has_broken_spirv_clamp = true,
           .has_broken_unsigned_image_offsets = true,

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -329,6 +329,11 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
 
         .lower_left_origin_mode = false,
         .need_declared_frag_colors = false,
+        .need_gather_subpixel_offset = driver_id == VK_DRIVER_ID_AMD_PROPRIETARY ||
+                                       driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE ||
+                                       driver_id == VK_DRIVER_ID_MESA_RADV ||
+                                       driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS ||
+                                       driver_id == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA,
 
         .has_broken_spirv_clamp = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS,
         .has_broken_spirv_position_input = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY,

src/video_core/texture_cache/texture_cache.h

@@ -1176,13 +1176,13 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA
     const size_t size_bytes = CalculateGuestSizeInBytes(new_info);
     const bool broken_views = runtime.HasBrokenTextureViewFormats();
     const bool native_bgr = runtime.HasNativeBgr();
-    std::vector<ImageId> overlap_ids;
+    boost::container::small_vector<ImageId, 4> overlap_ids;
     std::unordered_set<ImageId> overlaps_found;
-    std::vector<ImageId> left_aliased_ids;
-    std::vector<ImageId> right_aliased_ids;
+    boost::container::small_vector<ImageId, 4> left_aliased_ids;
+    boost::container::small_vector<ImageId, 4> right_aliased_ids;
     std::unordered_set<ImageId> ignore_textures;
-    std::vector<ImageId> bad_overlap_ids;
-    std::vector<ImageId> all_siblings;
+    boost::container::small_vector<ImageId, 4> bad_overlap_ids;
+    boost::container::small_vector<ImageId, 4> all_siblings;
     const bool this_is_linear = info.type == ImageType::Linear;
     const auto region_check = [&](ImageId overlap_id, ImageBase& overlap) {
         if (True(overlap.flags & ImageFlagBits::Remapped)) {
@@ -1298,16 +1298,16 @@ ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VA
         Image& overlap = slot_images[overlap_id];
         if (True(overlap.flags & ImageFlagBits::GpuModified)) {
             new_image.flags |= ImageFlagBits::GpuModified;
-        }
-        const auto& resolution = Settings::values.resolution_info;
-        const SubresourceBase base = new_image.TryFindBase(overlap.gpu_addr).value();
-        const u32 up_scale = can_rescale ? resolution.up_scale : 1;
-        const u32 down_shift = can_rescale ? resolution.down_shift : 0;
-        auto copies = MakeShrinkImageCopies(new_info, overlap.info, base, up_scale, down_shift);
-        if (overlap.info.num_samples != new_image.info.num_samples) {
-            runtime.CopyImageMSAA(new_image, overlap, std::move(copies));
-        } else {
-            runtime.CopyImage(new_image, overlap, std::move(copies));
+            const auto& resolution = Settings::values.resolution_info;
+            const SubresourceBase base = new_image.TryFindBase(overlap.gpu_addr).value();
+            const u32 up_scale = can_rescale ? resolution.up_scale : 1;
+            const u32 down_shift = can_rescale ? resolution.down_shift : 0;
+            auto copies = MakeShrinkImageCopies(new_info, overlap.info, base, up_scale, down_shift);
+            if (overlap.info.num_samples != new_image.info.num_samples) {
+                runtime.CopyImageMSAA(new_image, overlap, std::move(copies));
+            } else {
+                runtime.CopyImage(new_image, overlap, std::move(copies));
+            }
         }
         if (True(overlap.flags & ImageFlagBits::Tracked)) {
             UntrackImage(overlap, overlap_id);

src/yuzu/applets/qt_controller.cpp

@@ -300,7 +300,7 @@ bool QtControllerSelectorDialog::CheckIfParametersMet() {
     if (num_connected_players < min_supported_players ||
         num_connected_players > max_supported_players) {
         parameters_met = false;
-        ui->buttonBox->setEnabled(parameters_met);
+        ui->buttonBox->button(QDialogButtonBox::Ok)->setEnabled(parameters_met);
         return parameters_met;
     }
 
@@ -327,7 +327,7 @@ bool QtControllerSelectorDialog::CheckIfParametersMet() {
     }();
 
     parameters_met = all_controllers_compatible;
-    ui->buttonBox->setEnabled(parameters_met);
+    ui->buttonBox->button(QDialogButtonBox::Ok)->setEnabled(parameters_met);
     return parameters_met;
 }
 
@@ -697,8 +697,8 @@ void QtControllerSelector::ReconfigureControllers(
     emit MainWindowReconfigureControllers(parameters);
 }
 
-void QtControllerSelector::MainWindowReconfigureFinished() {
+void QtControllerSelector::MainWindowReconfigureFinished(bool is_success) {
     if (callback) {
-        callback();
+        callback(is_success);
     }
 }

src/yuzu/applets/qt_controller.h

@@ -167,7 +167,7 @@ signals:
     void MainWindowRequestExit() const;
 
 private:
-    void MainWindowReconfigureFinished();
+    void MainWindowReconfigureFinished(bool is_success);
 
     mutable ReconfigureCallback callback;
 };

src/yuzu/applets/qt_controller.ui

@@ -2629,7 +2629,7 @@
             <bool>true</bool>
            </property>
            <property name="standardButtons">
-            <set>QDialogButtonBox::Ok</set>
+            <set>QDialogButtonBox::Cancel|QDialogButtonBox::Ok</set>
            </property>
           </widget>
          </item>
@@ -2649,5 +2649,11 @@
    <receiver>QtControllerSelectorDialog</receiver>
    <slot>accept()</slot>
   </connection>
+  <connection>
+   <sender>buttonBox</sender>
+   <signal>rejected()</signal>
+   <receiver>QtControllerSelectorDialog</receiver>
+   <slot>reject()</slot>
+  </connection>
  </connections>
 </ui>

src/yuzu/applets/qt_profile_select.cpp

@@ -46,11 +46,13 @@ QPixmap GetIcon(Common::UUID uuid) {
 }
 } // Anonymous namespace
 
-QtProfileSelectionDialog::QtProfileSelectionDialog(Core::HID::HIDCore& hid_core, QWidget* parent)
+QtProfileSelectionDialog::QtProfileSelectionDialog(
+    Core::HID::HIDCore& hid_core, QWidget* parent,
+    const Core::Frontend::ProfileSelectParameters& parameters)
     : QDialog(parent), profile_manager(std::make_unique<Service::Account::ProfileManager>()) {
     outer_layout = new QVBoxLayout;
 
-    instruction_label = new QLabel(tr("Select a user:"));
+    instruction_label = new QLabel();
 
     scroll_area = new QScrollArea;
 
@@ -120,7 +122,8 @@ QtProfileSelectionDialog::QtProfileSelectionDialog(Core::HID::HIDCore& hid_core,
         item_model->appendRow(item);
 
     setLayout(outer_layout);
-    setWindowTitle(tr("Profile Selector"));
+    SetWindowTitle(parameters);
+    SetDialogPurpose(parameters);
     resize(550, 400);
 }
 
@@ -154,6 +157,76 @@ void QtProfileSelectionDialog::SelectUser(const QModelIndex& index) {
     user_index = index.row();
 }
 
+void QtProfileSelectionDialog::SetWindowTitle(
+    const Core::Frontend::ProfileSelectParameters& parameters) {
+    using Service::AM::Applets::UiMode;
+    switch (parameters.mode) {
+    case UiMode::UserCreator:
+    case UiMode::UserCreatorForStarter:
+        setWindowTitle(tr("Profile Creator"));
+        return;
+    case UiMode::EnsureNetworkServiceAccountAvailable:
+        setWindowTitle(tr("Profile Selector"));
+        return;
+    case UiMode::UserIconEditor:
+        setWindowTitle(tr("Profile Icon Editor"));
+        return;
+    case UiMode::UserNicknameEditor:
+        setWindowTitle(tr("Profile Nickname Editor"));
+        return;
+    case UiMode::NintendoAccountAuthorizationRequestContext:
+    case UiMode::IntroduceExternalNetworkServiceAccount:
+    case UiMode::IntroduceExternalNetworkServiceAccountForRegistration:
+    case UiMode::NintendoAccountNnidLinker:
+    case UiMode::LicenseRequirementsForNetworkService:
+    case UiMode::LicenseRequirementsForNetworkServiceWithUserContextImpl:
+    case UiMode::UserCreatorForImmediateNaLoginTest:
+    case UiMode::UserQualificationPromoter:
+    case UiMode::UserSelector:
+    default:
+        setWindowTitle(tr("Profile Selector"));
+    }
+}
+
+void QtProfileSelectionDialog::SetDialogPurpose(
+    const Core::Frontend::ProfileSelectParameters& parameters) {
+    using Service::AM::Applets::UserSelectionPurpose;
+
+    switch (parameters.purpose) {
+    case UserSelectionPurpose::GameCardRegistration:
+        instruction_label->setText(tr("Who will receive the points?"));
+        return;
+    case UserSelectionPurpose::EShopLaunch:
+        instruction_label->setText(tr("Who is using Nintendo eShop?"));
+        return;
+    case UserSelectionPurpose::EShopItemShow:
+        instruction_label->setText(tr("Who is making this purchase?"));
+        return;
+    case UserSelectionPurpose::PicturePost:
+        instruction_label->setText(tr("Who is posting?"));
+        return;
+    case UserSelectionPurpose::NintendoAccountLinkage:
+        instruction_label->setText(tr("Select a user to link to a Nintendo Account."));
+        return;
+    case UserSelectionPurpose::SettingsUpdate:
+        instruction_label->setText(tr("Change settings for which user?"));
+        return;
+    case UserSelectionPurpose::SaveDataDeletion:
+        instruction_label->setText(tr("Format data for which user?"));
+        return;
+    case UserSelectionPurpose::UserMigration:
+        instruction_label->setText(tr("Which user will be transferred to another console?"));
+        return;
+    case UserSelectionPurpose::SaveDataTransfer:
+        instruction_label->setText(tr("Send save data for which user?"));
+        return;
+    case UserSelectionPurpose::General:
+    default:
+        instruction_label->setText(tr("Select a user:"));
+        return;
+    }
+}
+
 QtProfileSelector::QtProfileSelector(GMainWindow& parent) {
     connect(this, &QtProfileSelector::MainWindowSelectProfile, &parent,
             &GMainWindow::ProfileSelectorSelectProfile, Qt::QueuedConnection);
@@ -170,9 +243,11 @@ void QtProfileSelector::Close() const {
     emit MainWindowRequestExit();
 }
 
-void QtProfileSelector::SelectProfile(SelectProfileCallback callback_) const {
+void QtProfileSelector::SelectProfile(
+    SelectProfileCallback callback_,
+    const Core::Frontend::ProfileSelectParameters& parameters) const {
     callback = std::move(callback_);
-    emit MainWindowSelectProfile();
+    emit MainWindowSelectProfile(parameters);
 }
 
 void QtProfileSelector::MainWindowFinishedSelection(std::optional<Common::UUID> uuid) {

src/yuzu/applets/qt_profile_select.h

@@ -28,7 +28,8 @@ class QtProfileSelectionDialog final : public QDialog {
     Q_OBJECT
 
 public:
-    explicit QtProfileSelectionDialog(Core::HID::HIDCore& hid_core, QWidget* parent);
+    explicit QtProfileSelectionDialog(Core::HID::HIDCore& hid_core, QWidget* parent,
+                                      const Core::Frontend::ProfileSelectParameters& parameters);
     ~QtProfileSelectionDialog() override;
 
     int exec() override;
@@ -40,6 +41,9 @@ public:
 private:
     void SelectUser(const QModelIndex& index);
 
+    void SetWindowTitle(const Core::Frontend::ProfileSelectParameters& parameters);
+    void SetDialogPurpose(const Core::Frontend::ProfileSelectParameters& parameters);
+
     int user_index = 0;
 
     QVBoxLayout* layout;
@@ -66,10 +70,11 @@ public:
     ~QtProfileSelector() override;
 
     void Close() const override;
-    void SelectProfile(SelectProfileCallback callback_) const override;
+    void SelectProfile(SelectProfileCallback callback_,
+                       const Core::Frontend::ProfileSelectParameters& parameters) const override;
 
 signals:
-    void MainWindowSelectProfile() const;
+    void MainWindowSelectProfile(const Core::Frontend::ProfileSelectParameters& parameters) const;
     void MainWindowRequestExit() const;
 
 private:

src/yuzu/main.cpp

@@ -576,6 +576,10 @@ void GMainWindow::RegisterMetaTypes() {
     // Controller Applet
     qRegisterMetaType<Core::Frontend::ControllerParameters>("Core::Frontend::ControllerParameters");
 
+    // Profile Select Applet
+    qRegisterMetaType<Core::Frontend::ProfileSelectParameters>(
+        "Core::Frontend::ProfileSelectParameters");
+
     // Software Keyboard Applet
     qRegisterMetaType<Core::Frontend::KeyboardInitializeParameters>(
         "Core::Frontend::KeyboardInitializeParameters");
@@ -634,15 +638,16 @@ void GMainWindow::ControllerSelectorReconfigureControllers(
                                       Qt::WindowStaysOnTopHint | Qt::WindowTitleHint |
                                       Qt::WindowSystemMenuHint);
     controller_applet->setWindowModality(Qt::WindowModal);
-    controller_applet->exec();
-
-    emit ControllerSelectorReconfigureFinished();
+    bool is_success = controller_applet->exec() != QDialog::Rejected;
 
     // Don't forget to apply settings.
+    system->HIDCore().DisableAllControllerConfiguration();
     system->ApplySettings();
     config->Save();
 
     UpdateStatusButtons();
+
+    emit ControllerSelectorReconfigureFinished(is_success);
 }
 
 void GMainWindow::ControllerSelectorRequestExit() {
@@ -651,8 +656,9 @@ void GMainWindow::ControllerSelectorRequestExit() {
     }
 }
 
-void GMainWindow::ProfileSelectorSelectProfile() {
-    profile_select_applet = new QtProfileSelectionDialog(system->HIDCore(), this);
+void GMainWindow::ProfileSelectorSelectProfile(
+    const Core::Frontend::ProfileSelectParameters& parameters) {
+    profile_select_applet = new QtProfileSelectionDialog(system->HIDCore(), this, parameters);
     SCOPE_EXIT({
         profile_select_applet->deleteLater();
         profile_select_applet = nullptr;
@@ -1719,8 +1725,9 @@ bool GMainWindow::LoadROM(const QString& filename, u64 program_id, std::size_t p
     return true;
 }
 
-bool GMainWindow::SelectAndSetCurrentUser() {
-    QtProfileSelectionDialog dialog(system->HIDCore(), this);
+bool GMainWindow::SelectAndSetCurrentUser(
+    const Core::Frontend::ProfileSelectParameters& parameters) {
+    QtProfileSelectionDialog dialog(system->HIDCore(), this, parameters);
     dialog.setWindowFlags(Qt::Dialog | Qt::CustomizeWindowHint | Qt::WindowTitleHint |
                           Qt::WindowSystemMenuHint | Qt::WindowCloseButtonHint);
     dialog.setWindowModality(Qt::WindowModal);
@@ -1766,7 +1773,13 @@ void GMainWindow::BootGame(const QString& filename, u64 program_id, std::size_t
     Settings::LogSettings();
 
     if (UISettings::values.select_user_on_boot) {
-        if (SelectAndSetCurrentUser() == false) {
+        const Core::Frontend::ProfileSelectParameters parameters{
+            .mode = Service::AM::Applets::UiMode::UserSelector,
+            .invalid_uid_list = {},
+            .display_options = {},
+            .purpose = Service::AM::Applets::UserSelectionPurpose::General,
+        };
+        if (SelectAndSetCurrentUser(parameters) == false) {
             return;
         }
     }
@@ -2058,7 +2071,13 @@ void GMainWindow::OnGameListOpenFolder(u64 program_id, GameListOpenTarget target
         if (has_user_save) {
             // User save data
             const auto select_profile = [this] {
-                QtProfileSelectionDialog dialog(system->HIDCore(), this);
+                const Core::Frontend::ProfileSelectParameters parameters{
+                    .mode = Service::AM::Applets::UiMode::UserSelector,
+                    .invalid_uid_list = {},
+                    .display_options = {},
+                    .purpose = Service::AM::Applets::UserSelectionPurpose::General,
+                };
+                QtProfileSelectionDialog dialog(system->HIDCore(), this, parameters);
                 dialog.setWindowFlags(Qt::Dialog | Qt::CustomizeWindowHint | Qt::WindowTitleHint |
                                       Qt::WindowSystemMenuHint | Qt::WindowCloseButtonHint);
                 dialog.setWindowModality(Qt::WindowModal);

src/yuzu/main.h

@@ -69,6 +69,7 @@ struct ControllerParameters;
 struct InlineAppearParameters;
 struct InlineTextParameters;
 struct KeyboardInitializeParameters;
+struct ProfileSelectParameters;
 } // namespace Core::Frontend
 
 namespace DiscordRPC {
@@ -165,7 +166,7 @@ signals:
 
     void AmiiboSettingsFinished(bool is_success, const std::string& name);
 
-    void ControllerSelectorReconfigureFinished();
+    void ControllerSelectorReconfigureFinished(bool is_success);
 
     void ErrorDisplayFinished();
 
@@ -203,7 +204,7 @@ public slots:
     void SoftwareKeyboardExit();
     void ErrorDisplayDisplayError(QString error_code, QString error_text);
     void ErrorDisplayRequestExit();
-    void ProfileSelectorSelectProfile();
+    void ProfileSelectorSelectProfile(const Core::Frontend::ProfileSelectParameters& parameters);
     void ProfileSelectorRequestExit();
     void WebBrowserOpenWebPage(const std::string& main_url, const std::string& additional_args,
                                bool is_local);
@@ -242,7 +243,7 @@ private:
     void SetDiscordEnabled(bool state);
     void LoadAmiibo(const QString& filename);
 
-    bool SelectAndSetCurrentUser();
+    bool SelectAndSetCurrentUser(const Core::Frontend::ProfileSelectParameters& parameters);
 
     /**
      * Stores the filename in the recently loaded files list.