fix(dist): wrap screenshots in <image> tags

The [appstream] spec says that <screenshot/> tags must be wrapped in
either <image/> or <video/> tags, so this patch does just that.

[appstream]: https://freedesktop.org/software/appstream/docs/chap-Metadata.html#tag-screenshots

dist/org.yuzu_emu.yuzu.metainfo.xml

@@ -53,10 +53,10 @@ SPDX-License-Identifier: CC0-1.0
   <developer_name>yuzu Emulator Team</developer_name>
   <content_rating type="oars-1.0"/>
   <screenshots>
-    <screenshot type="default">https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/001-Super%20Mario%20Odyssey%20.png</screenshot>
-    <screenshot>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/004-The%20Legend%20of%20Zelda%20Skyward%20Sword%20HD.png</screenshot>
-    <screenshot>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/007-Pokemon%20Sword.png</screenshot>
-    <screenshot>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/010-Hyrule%20Warriors%20Age%20of%20Calamity.png</screenshot>
-    <screenshot>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/039-Pok%C3%A9mon%20Mystery%20Dungeon%20Rescue%20Team%20DX.png.png.png</screenshot>
+    <screenshot type="default"><image>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/001-Super%20Mario%20Odyssey%20.png</image></screenshot>
+    <screenshot><image>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/004-The%20Legend%20of%20Zelda%20Skyward%20Sword%20HD.png</image></screenshot>
+    <screenshot><image>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/007-Pokemon%20Sword.png</image></screenshot>
+    <screenshot><image>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/010-Hyrule%20Warriors%20Age%20of%20Calamity.png</image></screenshot>
+    <screenshot><image>https://raw.githubusercontent.com/yuzu-emu/yuzu-emu.github.io/master/images/screenshots/039-Pok%C3%A9mon%20Mystery%20Dungeon%20Rescue%20Team%20DX.png.png.png</image></screenshot>
   </screenshots>
 </component>

src/audio_core/renderer/command/resample/upsample.cpp

@@ -20,25 +20,25 @@ static void SrcProcessFrame(std::span<s32> output, std::span<const s32> input,
                             const u32 target_sample_count, const u32 source_sample_count,
                             UpsamplerState* state) {
     constexpr u32 WindowSize = 10;
-    constexpr std::array<Common::FixedPoint<24, 8>, WindowSize> SincWindow1{
-        51.93359375f, -18.80078125f, 9.73046875f, -5.33203125f, 2.84375f,
-        -1.41015625f, 0.62109375f,   -0.2265625f, 0.0625f,      -0.00390625f,
+    constexpr std::array<Common::FixedPoint<17, 15>, WindowSize> WindowedSinc1{
+        0.95376587f,   -0.12872314f, 0.060028076f,  -0.032470703f, 0.017669678f,
+        -0.009124756f, 0.004272461f, -0.001739502f, 0.000579834f,  -0.000091552734f,
     };
-    constexpr std::array<Common::FixedPoint<24, 8>, WindowSize> SincWindow2{
-        105.35546875f, -24.52734375f, 11.9609375f,  -6.515625f, 3.52734375f,
-        -1.796875f,    0.828125f,     -0.32421875f, 0.1015625f, -0.015625f,
+    constexpr std::array<Common::FixedPoint<17, 15>, WindowSize> WindowedSinc2{
+        0.8230896f,    -0.19161987f,  0.093444824f,  -0.05090332f,   0.027557373f,
+        -0.014038086f, 0.0064697266f, -0.002532959f, 0.00079345703f, -0.00012207031f,
     };
-    constexpr std::array<Common::FixedPoint<24, 8>, WindowSize> SincWindow3{
-        122.08203125f, -16.47656250f, 7.68359375f,  -4.15625000f, 2.26171875f,
-        -1.16796875f,  0.54687500f,   -0.22265625f, 0.07421875f,  -0.01171875f,
+    constexpr std::array<Common::FixedPoint<17, 15>, WindowSize> WindowedSinc3{
+        0.6298828f,    -0.19274902f, 0.09725952f,    -0.05319214f,  0.028625488f,
+        -0.014373779f, 0.006500244f, -0.0024719238f, 0.0007324219f, -0.000091552734f,
     };
-    constexpr std::array<Common::FixedPoint<24, 8>, WindowSize> SincWindow4{
-        23.73437500f, -9.62109375f, 5.07812500f,  -2.78125000f, 1.46875000f,
-        -0.71484375f, 0.30859375f,  -0.10546875f, 0.02734375f,  0.00000000f,
+    constexpr std::array<Common::FixedPoint<17, 15>, WindowSize> WindowedSinc4{
+        0.4057312f,    -0.1468811f,  0.07601929f,    -0.041656494f,  0.022216797f,
+        -0.011016846f, 0.004852295f, -0.0017700195f, 0.00048828125f, -0.000030517578f,
     };
-    constexpr std::array<Common::FixedPoint<24, 8>, WindowSize> SincWindow5{
-        80.62500000f, -24.67187500f, 12.44921875f, -6.80859375f, 3.66406250f,
-        -1.83984375f, 0.83203125f,   -0.31640625f, 0.09375000f,  -0.01171875f,
+    constexpr std::array<Common::FixedPoint<17, 15>, WindowSize> WindowedSinc5{
+        0.1854248f,    -0.075164795f, 0.03967285f,    -0.021728516f,  0.011474609f,
+        -0.005584717f, 0.0024108887f, -0.0008239746f, 0.00021362305f, 0.0f,
     };
 
     if (!state->initialized) {
@@ -91,52 +91,31 @@ static void SrcProcessFrame(std::span<s32> output, std::span<const s32> input,
             static_cast<u16>((state->history_output_index + 1) % UpsamplerState::HistorySize);
     };
 
-    auto calculate_sample = [&state](std::span<const Common::FixedPoint<24, 8>> coeffs1,
-                                     std::span<const Common::FixedPoint<24, 8>> coeffs2) -> s32 {
+    auto calculate_sample = [&state](std::span<const Common::FixedPoint<17, 15>> coeffs1,
+                                     std::span<const Common::FixedPoint<17, 15>> coeffs2) -> s32 {
         auto output_index{state->history_output_index};
-        auto start_pos{output_index - state->history_start_index + 1U};
-        auto end_pos{10U};
+        u64 result{0};
 
-        if (start_pos < 10) {
-            end_pos = start_pos;
-        }
+        for (u32 coeff_index = 0; coeff_index < 10; coeff_index++) {
+            result += static_cast<u64>(state->history[output_index].to_raw()) *
+                      coeffs1[coeff_index].to_raw();
 
-        u64 prev_contrib{0};
-        u32 coeff_index{0};
-        for (; coeff_index < end_pos; coeff_index++, output_index--) {
-            prev_contrib += static_cast<u64>(state->history[output_index].to_raw()) *
-                            coeffs1[coeff_index].to_raw();
-        }
-
-        auto end_index{state->history_end_index};
-        for (; start_pos < 9; start_pos++, coeff_index++, end_index--) {
-            prev_contrib += static_cast<u64>(state->history[end_index].to_raw()) *
-                            coeffs1[coeff_index].to_raw();
+            output_index = output_index == state->history_start_index ? state->history_end_index
+                                                                      : output_index - 1;
         }
 
         output_index =
             static_cast<u16>((state->history_output_index + 1) % UpsamplerState::HistorySize);
-        start_pos = state->history_end_index - output_index + 1U;
-        end_pos = 10U;
 
-        if (start_pos < 10) {
-            end_pos = start_pos;
+        for (u32 coeff_index = 0; coeff_index < 10; coeff_index++) {
+            result += static_cast<u64>(state->history[output_index].to_raw()) *
+                      coeffs2[coeff_index].to_raw();
+
+            output_index = output_index == state->history_end_index ? state->history_start_index
+                                                                    : output_index + 1;
         }
 
-        u64 next_contrib{0};
-        coeff_index = 0;
-        for (; coeff_index < end_pos; coeff_index++, output_index++) {
-            next_contrib += static_cast<u64>(state->history[output_index].to_raw()) *
-                            coeffs2[coeff_index].to_raw();
-        }
-
-        auto start_index{state->history_start_index};
-        for (; start_pos < 9; start_pos++, start_index++, coeff_index++) {
-            next_contrib += static_cast<u64>(state->history[start_index].to_raw()) *
-                            coeffs2[coeff_index].to_raw();
-        }
-
-        return static_cast<s32>(((prev_contrib >> 15) + (next_contrib >> 15)) >> 8);
+        return static_cast<s32>(result >> (8 + 15));
     };
 
     switch (state->ratio.to_int_floor()) {
@@ -150,23 +129,23 @@ static void SrcProcessFrame(std::span<s32> output, std::span<const s32> input,
                 break;
 
             case 1:
-                output[write_index] = calculate_sample(SincWindow3, SincWindow4);
+                output[write_index] = calculate_sample(WindowedSinc1, WindowedSinc5);
                 break;
 
             case 2:
-                output[write_index] = calculate_sample(SincWindow2, SincWindow1);
+                output[write_index] = calculate_sample(WindowedSinc2, WindowedSinc4);
                 break;
 
             case 3:
-                output[write_index] = calculate_sample(SincWindow5, SincWindow5);
+                output[write_index] = calculate_sample(WindowedSinc3, WindowedSinc3);
                 break;
 
             case 4:
-                output[write_index] = calculate_sample(SincWindow1, SincWindow2);
+                output[write_index] = calculate_sample(WindowedSinc4, WindowedSinc2);
                 break;
 
             case 5:
-                output[write_index] = calculate_sample(SincWindow4, SincWindow3);
+                output[write_index] = calculate_sample(WindowedSinc5, WindowedSinc1);
                 break;
             }
             state->sample_index = static_cast<u8>((state->sample_index + 1) % 6);
@@ -183,11 +162,11 @@ static void SrcProcessFrame(std::span<s32> output, std::span<const s32> input,
                 break;
 
             case 1:
-                output[write_index] = calculate_sample(SincWindow2, SincWindow1);
+                output[write_index] = calculate_sample(WindowedSinc2, WindowedSinc4);
                 break;
 
             case 2:
-                output[write_index] = calculate_sample(SincWindow1, SincWindow2);
+                output[write_index] = calculate_sample(WindowedSinc4, WindowedSinc2);
                 break;
             }
             state->sample_index = static_cast<u8>((state->sample_index + 1) % 3);
@@ -204,12 +183,12 @@ static void SrcProcessFrame(std::span<s32> output, std::span<const s32> input,
                 break;
 
             case 1:
-                output[write_index] = calculate_sample(SincWindow1, SincWindow2);
+                output[write_index] = calculate_sample(WindowedSinc4, WindowedSinc2);
                 break;
 
             case 2:
                 increment();
-                output[write_index] = calculate_sample(SincWindow2, SincWindow1);
+                output[write_index] = calculate_sample(WindowedSinc2, WindowedSinc4);
                 break;
             }
             state->sample_index = static_cast<u8>((state->sample_index + 1) % 3);

src/common/input.h

@@ -292,9 +292,6 @@ class InputDevice {
 public:
     virtual ~InputDevice() = default;
 
-    // Request input device to update if necessary
-    virtual void SoftUpdate() {}
-
     // Force input device to update data regardless of the current state
     virtual void ForceUpdate() {}
 

src/common/settings.cpp

@@ -129,6 +129,10 @@ void UpdateRescalingInfo() {
         info.up_scale = 1;
         info.down_shift = 0;
         break;
+    case ResolutionSetup::Res3_2X:
+        info.up_scale = 3;
+        info.down_shift = 1;
+        break;
     case ResolutionSetup::Res2X:
         info.up_scale = 2;
         info.down_shift = 0;
@@ -149,6 +153,14 @@ void UpdateRescalingInfo() {
         info.up_scale = 6;
         info.down_shift = 0;
         break;
+    case ResolutionSetup::Res7X:
+        info.up_scale = 7;
+        info.down_shift = 0;
+        break;
+    case ResolutionSetup::Res8X:
+        info.up_scale = 8;
+        info.down_shift = 0;
+        break;
     default:
         ASSERT(false);
         info.up_scale = 1;
@@ -185,6 +197,7 @@ void RestoreGlobalState(bool is_powered_on) {
     // Renderer
     values.fsr_sharpening_slider.SetGlobal(true);
     values.renderer_backend.SetGlobal(true);
+    values.renderer_force_max_clock.SetGlobal(true);
     values.vulkan_device.SetGlobal(true);
     values.aspect_ratio.SetGlobal(true);
     values.max_anisotropy.SetGlobal(true);

src/common/settings.h

@@ -56,11 +56,14 @@ enum class ResolutionSetup : u32 {
     Res1_2X = 0,
     Res3_4X = 1,
     Res1X = 2,
-    Res2X = 3,
-    Res3X = 4,
-    Res4X = 5,
-    Res5X = 6,
-    Res6X = 7,
+    Res3_2X = 3,
+    Res2X = 4,
+    Res3X = 5,
+    Res4X = 6,
+    Res5X = 7,
+    Res6X = 8,
+    Res7X = 9,
+    Res8X = 10,
 };
 
 enum class ScalingFilter : u32 {
@@ -415,6 +418,7 @@ struct Values {
     // Renderer
     SwitchableSetting<RendererBackend, true> renderer_backend{
         RendererBackend::Vulkan, RendererBackend::OpenGL, RendererBackend::Null, "backend"};
+    SwitchableSetting<bool> renderer_force_max_clock{false, "force_max_clock"};
     Setting<bool> renderer_debug{false, "debug"};
     Setting<bool> renderer_shader_feedback{false, "shader_feedback"};
     Setting<bool> enable_nsight_aftermath{false, "nsight_aftermath"};

src/core/CMakeLists.txt

@@ -226,6 +226,7 @@ add_library(core STATIC
     hle/kernel/k_page_buffer.h
     hle/kernel/k_page_heap.cpp
     hle/kernel/k_page_heap.h
+    hle/kernel/k_page_group.cpp
     hle/kernel/k_page_group.h
     hle/kernel/k_page_table.cpp
     hle/kernel/k_page_table.h

src/core/core_timing.cpp

@@ -142,16 +142,24 @@ void CoreTiming::ScheduleLoopingEvent(std::chrono::nanoseconds start_time,
 }
 
 void CoreTiming::UnscheduleEvent(const std::shared_ptr<EventType>& event_type,
-                                 std::uintptr_t user_data) {
-    std::scoped_lock scope{basic_lock};
-    const auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
-        return e.type.lock().get() == event_type.get() && e.user_data == user_data;
-    });
+                                 std::uintptr_t user_data, bool wait) {
+    {
+        std::scoped_lock lk{basic_lock};
+        const auto itr =
+            std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
+                return e.type.lock().get() == event_type.get() && e.user_data == user_data;
+            });
 
-    // Removing random items breaks the invariant so we have to re-establish it.
-    if (itr != event_queue.end()) {
-        event_queue.erase(itr, event_queue.end());
-        std::make_heap(event_queue.begin(), event_queue.end(), std::greater<>());
+        // Removing random items breaks the invariant so we have to re-establish it.
+        if (itr != event_queue.end()) {
+            event_queue.erase(itr, event_queue.end());
+            std::make_heap(event_queue.begin(), event_queue.end(), std::greater<>());
+        }
+    }
+
+    // Force any in-progress events to finish
+    if (wait) {
+        std::scoped_lock lk{advance_lock};
     }
 }
 
@@ -190,20 +198,6 @@ u64 CoreTiming::GetClockTicks() const {
     return CpuCyclesToClockCycles(ticks);
 }
 
-void CoreTiming::RemoveEvent(const std::shared_ptr<EventType>& event_type) {
-    std::scoped_lock lock{basic_lock};
-
-    const auto itr = std::remove_if(event_queue.begin(), event_queue.end(), [&](const Event& e) {
-        return e.type.lock().get() == event_type.get();
-    });
-
-    // Removing random items breaks the invariant so we have to re-establish it.
-    if (itr != event_queue.end()) {
-        event_queue.erase(itr, event_queue.end());
-        std::make_heap(event_queue.begin(), event_queue.end(), std::greater<>());
-    }
-}
-
 std::optional<s64> CoreTiming::Advance() {
     std::scoped_lock lock{advance_lock, basic_lock};
     global_timer = GetGlobalTimeNs().count();

src/core/core_timing.h

@@ -98,10 +98,13 @@ public:
                               const std::shared_ptr<EventType>& event_type,
                               std::uintptr_t user_data = 0, bool absolute_time = false);
 
-    void UnscheduleEvent(const std::shared_ptr<EventType>& event_type, std::uintptr_t user_data);
+    void UnscheduleEvent(const std::shared_ptr<EventType>& event_type, std::uintptr_t user_data,
+                         bool wait = true);
 
-    /// We only permit one event of each type in the queue at a time.
-    void RemoveEvent(const std::shared_ptr<EventType>& event_type);
+    void UnscheduleEventWithoutWait(const std::shared_ptr<EventType>& event_type,
+                                    std::uintptr_t user_data) {
+        UnscheduleEvent(event_type, user_data, false);
+    }
 
     void AddTicks(u64 ticks_to_add);
 

src/core/hid/emulated_controller.cpp

@@ -1434,16 +1434,6 @@ AnalogSticks EmulatedController::GetSticks() const {
         return {};
     }
 
-    // Some drivers like stick from buttons need constant refreshing
-    for (auto& device : stick_devices) {
-        if (!device) {
-            continue;
-        }
-        lock.unlock();
-        device->SoftUpdate();
-        lock.lock();
-    }
-
     return controller.analog_stick_state;
 }
 

src/core/hle/kernel/k_code_memory.cpp

@@ -27,13 +27,13 @@ Result KCodeMemory::Initialize(Core::DeviceMemory& device_memory, VAddr addr, si
     auto& page_table = m_owner->PageTable();
 
     // Construct the page group.
-    m_page_group = {};
+    m_page_group.emplace(kernel, page_table.GetBlockInfoManager());
 
     // Lock the memory.
-    R_TRY(page_table.LockForCodeMemory(&m_page_group, addr, size))
+    R_TRY(page_table.LockForCodeMemory(std::addressof(*m_page_group), addr, size))
 
     // Clear the memory.
-    for (const auto& block : m_page_group.Nodes()) {
+    for (const auto& block : *m_page_group) {
         std::memset(device_memory.GetPointer<void>(block.GetAddress()), 0xFF, block.GetSize());
     }
 
@@ -51,12 +51,13 @@ Result KCodeMemory::Initialize(Core::DeviceMemory& device_memory, VAddr addr, si
 void KCodeMemory::Finalize() {
     // Unlock.
     if (!m_is_mapped && !m_is_owner_mapped) {
-        const size_t size = m_page_group.GetNumPages() * PageSize;
-        m_owner->PageTable().UnlockForCodeMemory(m_address, size, m_page_group);
+        const size_t size = m_page_group->GetNumPages() * PageSize;
+        m_owner->PageTable().UnlockForCodeMemory(m_address, size, *m_page_group);
     }
 
     // Close the page group.
-    m_page_group = {};
+    m_page_group->Close();
+    m_page_group->Finalize();
 
     // Close our reference to our owner.
     m_owner->Close();
@@ -64,7 +65,7 @@ void KCodeMemory::Finalize() {
 
 Result KCodeMemory::Map(VAddr address, size_t size) {
     // Validate the size.
-    R_UNLESS(m_page_group.GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
+    R_UNLESS(m_page_group->GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
 
     // Lock ourselves.
     KScopedLightLock lk(m_lock);
@@ -74,7 +75,7 @@ Result KCodeMemory::Map(VAddr address, size_t size) {
 
     // Map the memory.
     R_TRY(kernel.CurrentProcess()->PageTable().MapPages(
-        address, m_page_group, KMemoryState::CodeOut, KMemoryPermission::UserReadWrite));
+        address, *m_page_group, KMemoryState::CodeOut, KMemoryPermission::UserReadWrite));
 
     // Mark ourselves as mapped.
     m_is_mapped = true;
@@ -84,13 +85,13 @@ Result KCodeMemory::Map(VAddr address, size_t size) {
 
 Result KCodeMemory::Unmap(VAddr address, size_t size) {
     // Validate the size.
-    R_UNLESS(m_page_group.GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
+    R_UNLESS(m_page_group->GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
 
     // Lock ourselves.
     KScopedLightLock lk(m_lock);
 
     // Unmap the memory.
-    R_TRY(kernel.CurrentProcess()->PageTable().UnmapPages(address, m_page_group,
+    R_TRY(kernel.CurrentProcess()->PageTable().UnmapPages(address, *m_page_group,
                                                           KMemoryState::CodeOut));
 
     // Mark ourselves as unmapped.
@@ -101,7 +102,7 @@ Result KCodeMemory::Unmap(VAddr address, size_t size) {
 
 Result KCodeMemory::MapToOwner(VAddr address, size_t size, Svc::MemoryPermission perm) {
     // Validate the size.
-    R_UNLESS(m_page_group.GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
+    R_UNLESS(m_page_group->GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
 
     // Lock ourselves.
     KScopedLightLock lk(m_lock);
@@ -125,7 +126,7 @@ Result KCodeMemory::MapToOwner(VAddr address, size_t size, Svc::MemoryPermission
 
     // Map the memory.
     R_TRY(
-        m_owner->PageTable().MapPages(address, m_page_group, KMemoryState::GeneratedCode, k_perm));
+        m_owner->PageTable().MapPages(address, *m_page_group, KMemoryState::GeneratedCode, k_perm));
 
     // Mark ourselves as mapped.
     m_is_owner_mapped = true;
@@ -135,13 +136,13 @@ Result KCodeMemory::MapToOwner(VAddr address, size_t size, Svc::MemoryPermission
 
 Result KCodeMemory::UnmapFromOwner(VAddr address, size_t size) {
     // Validate the size.
-    R_UNLESS(m_page_group.GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
+    R_UNLESS(m_page_group->GetNumPages() == Common::DivideUp(size, PageSize), ResultInvalidSize);
 
     // Lock ourselves.
     KScopedLightLock lk(m_lock);
 
     // Unmap the memory.
-    R_TRY(m_owner->PageTable().UnmapPages(address, m_page_group, KMemoryState::GeneratedCode));
+    R_TRY(m_owner->PageTable().UnmapPages(address, *m_page_group, KMemoryState::GeneratedCode));
 
     // Mark ourselves as unmapped.
     m_is_owner_mapped = false;

src/core/hle/kernel/k_code_memory.h

@@ -3,6 +3,8 @@
 
 #pragma once
 
+#include <optional>
+
 #include "common/common_types.h"
 #include "core/device_memory.h"
 #include "core/hle/kernel/k_auto_object.h"
@@ -49,11 +51,11 @@ public:
         return m_address;
     }
     size_t GetSize() const {
-        return m_is_initialized ? m_page_group.GetNumPages() * PageSize : 0;
+        return m_is_initialized ? m_page_group->GetNumPages() * PageSize : 0;
     }
 
 private:
-    KPageGroup m_page_group{};
+    std::optional<KPageGroup> m_page_group{};
     KProcess* m_owner{};
     VAddr m_address{};
     KLightLock m_lock;

src/core/hle/kernel/k_hardware_timer.cpp

@@ -18,7 +18,8 @@ void KHardwareTimer::Initialize() {
 }
 
 void KHardwareTimer::Finalize() {
-    this->DisableInterrupt();
+    m_kernel.System().CoreTiming().UnscheduleEvent(m_event_type, reinterpret_cast<uintptr_t>(this));
+    m_wakeup_time = std::numeric_limits<s64>::max();
     m_event_type.reset();
 }
 
@@ -59,7 +60,8 @@ void KHardwareTimer::EnableInterrupt(s64 wakeup_time) {
 }
 
 void KHardwareTimer::DisableInterrupt() {
-    m_kernel.System().CoreTiming().UnscheduleEvent(m_event_type, reinterpret_cast<uintptr_t>(this));
+    m_kernel.System().CoreTiming().UnscheduleEventWithoutWait(m_event_type,
+                                                              reinterpret_cast<uintptr_t>(this));
     m_wakeup_time = std::numeric_limits<s64>::max();
 }
 

src/core/hle/kernel/k_memory_manager.cpp

@@ -223,7 +223,7 @@ Result KMemoryManager::AllocatePageGroupImpl(KPageGroup* out, size_t num_pages,
 
     // Ensure that we don't leave anything un-freed.
     ON_RESULT_FAILURE {
-        for (const auto& it : out->Nodes()) {
+        for (const auto& it : *out) {
             auto& manager = this->GetManager(it.GetAddress());
             const size_t node_num_pages = std::min<u64>(
                 it.GetNumPages(), (manager.GetEndAddress() - it.GetAddress()) / PageSize);
@@ -285,7 +285,7 @@ Result KMemoryManager::AllocateAndOpen(KPageGroup* out, size_t num_pages, u32 op
                                       m_has_optimized_process[static_cast<size_t>(pool)], true));
 
     // Open the first reference to the pages.
-    for (const auto& block : out->Nodes()) {
+    for (const auto& block : *out) {
         PAddr cur_address = block.GetAddress();
         size_t remaining_pages = block.GetNumPages();
         while (remaining_pages > 0) {
@@ -335,7 +335,7 @@ Result KMemoryManager::AllocateForProcess(KPageGroup* out, size_t num_pages, u32
     // Perform optimized memory tracking, if we should.
     if (optimized) {
         // Iterate over the allocated blocks.
-        for (const auto& block : out->Nodes()) {
+        for (const auto& block : *out) {
             // Get the block extents.
             const PAddr block_address = block.GetAddress();
             const size_t block_pages = block.GetNumPages();
@@ -391,7 +391,7 @@ Result KMemoryManager::AllocateForProcess(KPageGroup* out, size_t num_pages, u32
         }
     } else {
         // Set all the allocated memory.
-        for (const auto& block : out->Nodes()) {
+        for (const auto& block : *out) {
             std::memset(m_system.DeviceMemory().GetPointer<void>(block.GetAddress()), fill_pattern,
                         block.GetSize());
         }

src/core/hle/kernel/k_page_group.cpp

@@ -0,0 +1,121 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "core/hle/kernel/k_dynamic_resource_manager.h"
+#include "core/hle/kernel/k_memory_manager.h"
+#include "core/hle/kernel/k_page_group.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/svc_results.h"
+
+namespace Kernel {
+
+void KPageGroup::Finalize() {
+    KBlockInfo* cur = m_first_block;
+    while (cur != nullptr) {
+        KBlockInfo* next = cur->GetNext();
+        m_manager->Free(cur);
+        cur = next;
+    }
+
+    m_first_block = nullptr;
+    m_last_block = nullptr;
+}
+
+void KPageGroup::CloseAndReset() {
+    auto& mm = m_kernel.MemoryManager();
+
+    KBlockInfo* cur = m_first_block;
+    while (cur != nullptr) {
+        KBlockInfo* next = cur->GetNext();
+        mm.Close(cur->GetAddress(), cur->GetNumPages());
+        m_manager->Free(cur);
+        cur = next;
+    }
+
+    m_first_block = nullptr;
+    m_last_block = nullptr;
+}
+
+size_t KPageGroup::GetNumPages() const {
+    size_t num_pages = 0;
+
+    for (const auto& it : *this) {
+        num_pages += it.GetNumPages();
+    }
+
+    return num_pages;
+}
+
+Result KPageGroup::AddBlock(KPhysicalAddress addr, size_t num_pages) {
+    // Succeed immediately if we're adding no pages.
+    R_SUCCEED_IF(num_pages == 0);
+
+    // Check for overflow.
+    ASSERT(addr < addr + num_pages * PageSize);
+
+    // Try to just append to the last block.
+    if (m_last_block != nullptr) {
+        R_SUCCEED_IF(m_last_block->TryConcatenate(addr, num_pages));
+    }
+
+    // Allocate a new block.
+    KBlockInfo* new_block = m_manager->Allocate();
+    R_UNLESS(new_block != nullptr, ResultOutOfResource);
+
+    // Initialize the block.
+    new_block->Initialize(addr, num_pages);
+
+    // Add the block to our list.
+    if (m_last_block != nullptr) {
+        m_last_block->SetNext(new_block);
+    } else {
+        m_first_block = new_block;
+    }
+    m_last_block = new_block;
+
+    R_SUCCEED();
+}
+
+void KPageGroup::Open() const {
+    auto& mm = m_kernel.MemoryManager();
+
+    for (const auto& it : *this) {
+        mm.Open(it.GetAddress(), it.GetNumPages());
+    }
+}
+
+void KPageGroup::OpenFirst() const {
+    auto& mm = m_kernel.MemoryManager();
+
+    for (const auto& it : *this) {
+        mm.OpenFirst(it.GetAddress(), it.GetNumPages());
+    }
+}
+
+void KPageGroup::Close() const {
+    auto& mm = m_kernel.MemoryManager();
+
+    for (const auto& it : *this) {
+        mm.Close(it.GetAddress(), it.GetNumPages());
+    }
+}
+
+bool KPageGroup::IsEquivalentTo(const KPageGroup& rhs) const {
+    auto lit = this->begin();
+    auto rit = rhs.begin();
+    auto lend = this->end();
+    auto rend = rhs.end();
+
+    while (lit != lend && rit != rend) {
+        if (*lit != *rit) {
+            return false;
+        }
+
+        ++lit;
+        ++rit;
+    }
+
+    return lit == lend && rit == rend;
+}
+
+} // namespace Kernel

src/core/hle/kernel/k_page_group.h

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #pragma once
@@ -13,24 +13,23 @@
 
 namespace Kernel {
 
+class KBlockInfoManager;
+class KernelCore;
 class KPageGroup;
 
 class KBlockInfo {
-private:
-    friend class KPageGroup;
-
 public:
-    constexpr KBlockInfo() = default;
+    constexpr explicit KBlockInfo() : m_next(nullptr) {}
 
-    constexpr void Initialize(PAddr addr, size_t np) {
+    constexpr void Initialize(KPhysicalAddress addr, size_t np) {
         ASSERT(Common::IsAligned(addr, PageSize));
         ASSERT(static_cast<u32>(np) == np);
 
-        m_page_index = static_cast<u32>(addr) / PageSize;
+        m_page_index = static_cast<u32>(addr / PageSize);
         m_num_pages = static_cast<u32>(np);
     }
 
-    constexpr PAddr GetAddress() const {
+    constexpr KPhysicalAddress GetAddress() const {
         return m_page_index * PageSize;
     }
     constexpr size_t GetNumPages() const {
@@ -39,10 +38,10 @@ public:
     constexpr size_t GetSize() const {
         return this->GetNumPages() * PageSize;
     }
-    constexpr PAddr GetEndAddress() const {
+    constexpr KPhysicalAddress GetEndAddress() const {
         return (m_page_index + m_num_pages) * PageSize;
     }
-    constexpr PAddr GetLastAddress() const {
+    constexpr KPhysicalAddress GetLastAddress() const {
         return this->GetEndAddress() - 1;
     }
 
@@ -62,8 +61,8 @@ public:
         return !(*this == rhs);
     }
 
-    constexpr bool IsStrictlyBefore(PAddr addr) const {
-        const PAddr end = this->GetEndAddress();
+    constexpr bool IsStrictlyBefore(KPhysicalAddress addr) const {
+        const KPhysicalAddress end = this->GetEndAddress();
 
         if (m_page_index != 0 && end == 0) {
             return false;
@@ -72,11 +71,11 @@ public:
         return end < addr;
     }
 
-    constexpr bool operator<(PAddr addr) const {
+    constexpr bool operator<(KPhysicalAddress addr) const {
         return this->IsStrictlyBefore(addr);
     }
 
-    constexpr bool TryConcatenate(PAddr addr, size_t np) {
+    constexpr bool TryConcatenate(KPhysicalAddress addr, size_t np) {
         if (addr != 0 && addr == this->GetEndAddress()) {
             m_num_pages += static_cast<u32>(np);
             return true;
@@ -90,96 +89,118 @@ private:
     }
 
 private:
+    friend class KPageGroup;
+
     KBlockInfo* m_next{};
     u32 m_page_index{};
     u32 m_num_pages{};
 };
 static_assert(sizeof(KBlockInfo) <= 0x10);
 
-class KPageGroup final {
+class KPageGroup {
 public:
-    class Node final {
+    class Iterator {
     public:
-        constexpr Node(u64 addr_, std::size_t num_pages_) : addr{addr_}, num_pages{num_pages_} {}
+        using iterator_category = std::forward_iterator_tag;
+        using value_type = const KBlockInfo;
+        using difference_type = std::ptrdiff_t;
+        using pointer = value_type*;
+        using reference = value_type&;
 
-        constexpr u64 GetAddress() const {
-            return addr;
+        constexpr explicit Iterator(pointer n) : m_node(n) {}
+
+        constexpr bool operator==(const Iterator& rhs) const {
+            return m_node == rhs.m_node;
+        }
+        constexpr bool operator!=(const Iterator& rhs) const {
+            return !(*this == rhs);
         }
 
-        constexpr std::size_t GetNumPages() const {
-            return num_pages;
+        constexpr pointer operator->() const {
+            return m_node;
+        }
+        constexpr reference operator*() const {
+            return *m_node;
         }
 
-        constexpr std::size_t GetSize() const {
-            return GetNumPages() * PageSize;
+        constexpr Iterator& operator++() {
+            m_node = m_node->GetNext();
+            return *this;
+        }
+
+        constexpr Iterator operator++(int) {
+            const Iterator it{*this};
+            ++(*this);
+            return it;
         }
 
     private:
-        u64 addr{};
-        std::size_t num_pages{};
+        pointer m_node{};
     };
 
-public:
-    KPageGroup() = default;
-    KPageGroup(u64 address, u64 num_pages) {
-        ASSERT(AddBlock(address, num_pages).IsSuccess());
+    explicit KPageGroup(KernelCore& kernel, KBlockInfoManager* m)
+        : m_kernel{kernel}, m_manager{m} {}
+    ~KPageGroup() {
+        this->Finalize();
     }
 
-    constexpr std::list<Node>& Nodes() {
-        return nodes;
+    void CloseAndReset();
+    void Finalize();
+
+    Iterator begin() const {
+        return Iterator{m_first_block};
+    }
+    Iterator end() const {
+        return Iterator{nullptr};
+    }
+    bool empty() const {
+        return m_first_block == nullptr;
     }
 
-    constexpr const std::list<Node>& Nodes() const {
-        return nodes;
+    Result AddBlock(KPhysicalAddress addr, size_t num_pages);
+    void Open() const;
+    void OpenFirst() const;
+    void Close() const;
+
+    size_t GetNumPages() const;
+
+    bool IsEquivalentTo(const KPageGroup& rhs) const;
+
+    bool operator==(const KPageGroup& rhs) const {
+        return this->IsEquivalentTo(rhs);
     }
 
-    std::size_t GetNumPages() const {
-        std::size_t num_pages = 0;
-        for (const Node& node : nodes) {
-            num_pages += node.GetNumPages();
-        }
-        return num_pages;
+    bool operator!=(const KPageGroup& rhs) const {
+        return !(*this == rhs);
     }
 
-    bool IsEqual(KPageGroup& other) const {
-        auto this_node = nodes.begin();
-        auto other_node = other.nodes.begin();
-        while (this_node != nodes.end() && other_node != other.nodes.end()) {
-            if (this_node->GetAddress() != other_node->GetAddress() ||
-                this_node->GetNumPages() != other_node->GetNumPages()) {
-                return false;
-            }
-            this_node = std::next(this_node);
-            other_node = std::next(other_node);
-        }
-
-        return this_node == nodes.end() && other_node == other.nodes.end();
-    }
-
-    Result AddBlock(u64 address, u64 num_pages) {
-        if (!num_pages) {
-            return ResultSuccess;
-        }
-        if (!nodes.empty()) {
-            const auto node = nodes.back();
-            if (node.GetAddress() + node.GetNumPages() * PageSize == address) {
-                address = node.GetAddress();
-                num_pages += node.GetNumPages();
-                nodes.pop_back();
-            }
-        }
-        nodes.push_back({address, num_pages});
-        return ResultSuccess;
-    }
-
-    bool Empty() const {
-        return nodes.empty();
-    }
-
-    void Finalize() {}
-
 private:
-    std::list<Node> nodes;
+    KernelCore& m_kernel;
+    KBlockInfo* m_first_block{};
+    KBlockInfo* m_last_block{};
+    KBlockInfoManager* m_manager{};
+};
+
+class KScopedPageGroup {
+public:
+    explicit KScopedPageGroup(const KPageGroup* gp) : m_pg(gp) {
+        if (m_pg) {
+            m_pg->Open();
+        }
+    }
+    explicit KScopedPageGroup(const KPageGroup& gp) : KScopedPageGroup(std::addressof(gp)) {}
+    ~KScopedPageGroup() {
+        if (m_pg) {
+            m_pg->Close();
+        }
+    }
+
+    void CancelClose() {
+        m_pg = nullptr;
+    }
+
+private:
+    const KPageGroup* m_pg{};
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_page_table.cpp

@@ -100,7 +100,7 @@ constexpr size_t GetAddressSpaceWidthFromType(FileSys::ProgramAddressSpaceType a
 
 KPageTable::KPageTable(Core::System& system_)
     : m_general_lock{system_.Kernel()},
-      m_map_physical_memory_lock{system_.Kernel()}, m_system{system_} {}
+      m_map_physical_memory_lock{system_.Kernel()}, m_system{system_}, m_kernel{system_.Kernel()} {}
 
 KPageTable::~KPageTable() = default;
 
@@ -373,7 +373,7 @@ Result KPageTable::MapProcessCode(VAddr addr, size_t num_pages, KMemoryState sta
                                                  m_memory_block_slab_manager);
 
     // Allocate and open.
-    KPageGroup pg;
+    KPageGroup pg{m_kernel, m_block_info_manager};
     R_TRY(m_system.Kernel().MemoryManager().AllocateAndOpen(
         &pg, num_pages,
         KMemoryManager::EncodeOption(KMemoryManager::Pool::Application, m_allocation_option)));
@@ -432,7 +432,7 @@ Result KPageTable::MapCodeMemory(VAddr dst_address, VAddr src_address, size_t si
         const size_t num_pages = size / PageSize;
 
         // Create page groups for the memory being mapped.
-        KPageGroup pg;
+        KPageGroup pg{m_kernel, m_block_info_manager};
         AddRegionToPages(src_address, num_pages, pg);
 
         // Reprotect the source as kernel-read/not mapped.
@@ -593,7 +593,7 @@ Result KPageTable::MakePageGroup(KPageGroup& pg, VAddr addr, size_t num_pages) {
     const size_t size = num_pages * PageSize;
 
     // We're making a new group, not adding to an existing one.
-    R_UNLESS(pg.Empty(), ResultInvalidCurrentMemory);
+    R_UNLESS(pg.empty(), ResultInvalidCurrentMemory);
 
     // Begin traversal.
     Common::PageTable::TraversalContext context;
@@ -640,11 +640,10 @@ Result KPageTable::MakePageGroup(KPageGroup& pg, VAddr addr, size_t num_pages) {
     R_SUCCEED();
 }
 
-bool KPageTable::IsValidPageGroup(const KPageGroup& pg_ll, VAddr addr, size_t num_pages) {
+bool KPageTable::IsValidPageGroup(const KPageGroup& pg, VAddr addr, size_t num_pages) {
     ASSERT(this->IsLockedByCurrentThread());
 
     const size_t size = num_pages * PageSize;
-    const auto& pg = pg_ll.Nodes();
     const auto& memory_layout = m_system.Kernel().MemoryLayout();
 
     // Empty groups are necessarily invalid.
@@ -942,9 +941,6 @@ Result KPageTable::SetupForIpcServer(VAddr* out_addr, size_t size, VAddr src_add
 
     ON_RESULT_FAILURE {
         if (cur_mapped_addr != dst_addr) {
-            // HACK: Manually close the pages.
-            HACK_ClosePages(dst_addr, (cur_mapped_addr - dst_addr) / PageSize);
-
             ASSERT(Operate(dst_addr, (cur_mapped_addr - dst_addr) / PageSize,
                            KMemoryPermission::None, OperationType::Unmap)
                        .IsSuccess());
@@ -1020,9 +1016,6 @@ Result KPageTable::SetupForIpcServer(VAddr* out_addr, size_t size, VAddr src_add
         // Map the page.
         R_TRY(Operate(cur_mapped_addr, 1, test_perm, OperationType::Map, start_partial_page));
 
-        // HACK: Manually open the pages.
-        HACK_OpenPages(start_partial_page, 1);
-
         // Update tracking extents.
         cur_mapped_addr += PageSize;
         cur_block_addr += PageSize;
@@ -1051,9 +1044,6 @@ Result KPageTable::SetupForIpcServer(VAddr* out_addr, size_t size, VAddr src_add
             R_TRY(Operate(cur_mapped_addr, cur_block_size / PageSize, test_perm, OperationType::Map,
                           cur_block_addr));
 
-            // HACK: Manually open the pages.
-            HACK_OpenPages(cur_block_addr, cur_block_size / PageSize);
-
             // Update tracking extents.
             cur_mapped_addr += cur_block_size;
             cur_block_addr = next_entry.phys_addr;
@@ -1073,9 +1063,6 @@ Result KPageTable::SetupForIpcServer(VAddr* out_addr, size_t size, VAddr src_add
         R_TRY(Operate(cur_mapped_addr, last_block_size / PageSize, test_perm, OperationType::Map,
                       cur_block_addr));
 
-        // HACK: Manually open the pages.
-        HACK_OpenPages(cur_block_addr, last_block_size / PageSize);
-
         // Update tracking extents.
         cur_mapped_addr += last_block_size;
         cur_block_addr += last_block_size;
@@ -1107,9 +1094,6 @@ Result KPageTable::SetupForIpcServer(VAddr* out_addr, size_t size, VAddr src_add
 
         // Map the page.
         R_TRY(Operate(cur_mapped_addr, 1, test_perm, OperationType::Map, end_partial_page));
-
-        // HACK: Manually open the pages.
-        HACK_OpenPages(end_partial_page, 1);
     }
 
     // Update memory blocks to reflect our changes
@@ -1211,9 +1195,6 @@ Result KPageTable::CleanupForIpcServer(VAddr address, size_t size, KMemoryState
     const size_t aligned_size = aligned_end - aligned_start;
     const size_t aligned_num_pages = aligned_size / PageSize;
 
-    // HACK: Manually close the pages.
-    HACK_ClosePages(aligned_start, aligned_num_pages);
-
     // Unmap the pages.
     R_TRY(Operate(aligned_start, aligned_num_pages, KMemoryPermission::None, OperationType::Unmap));
 
@@ -1501,17 +1482,6 @@ void KPageTable::CleanupForIpcClientOnServerSetupFailure([[maybe_unused]] PageLi
     }
 }
 
-void KPageTable::HACK_OpenPages(PAddr phys_addr, size_t num_pages) {
-    m_system.Kernel().MemoryManager().OpenFirst(phys_addr, num_pages);
-}
-
-void KPageTable::HACK_ClosePages(VAddr virt_addr, size_t num_pages) {
-    for (size_t index = 0; index < num_pages; ++index) {
-        const auto paddr = GetPhysicalAddr(virt_addr + (index * PageSize));
-        m_system.Kernel().MemoryManager().Close(paddr, 1);
-    }
-}
-
 Result KPageTable::MapPhysicalMemory(VAddr address, size_t size) {
     // Lock the physical memory lock.
     KScopedLightLock phys_lk(m_map_physical_memory_lock);
@@ -1572,7 +1542,7 @@ Result KPageTable::MapPhysicalMemory(VAddr address, size_t size) {
             R_UNLESS(memory_reservation.Succeeded(), ResultLimitReached);
 
             // Allocate pages for the new memory.
-            KPageGroup pg;
+            KPageGroup pg{m_kernel, m_block_info_manager};
             R_TRY(m_system.Kernel().MemoryManager().AllocateForProcess(
                 &pg, (size - mapped_size) / PageSize, m_allocate_option, 0, 0));
 
@@ -1650,7 +1620,7 @@ Result KPageTable::MapPhysicalMemory(VAddr address, size_t size) {
                 KScopedPageTableUpdater updater(this);
 
                 // Prepare to iterate over the memory.
-                auto pg_it = pg.Nodes().begin();
+                auto pg_it = pg.begin();
                 PAddr pg_phys_addr = pg_it->GetAddress();
                 size_t pg_pages = pg_it->GetNumPages();
 
@@ -1680,9 +1650,6 @@ Result KPageTable::MapPhysicalMemory(VAddr address, size_t size) {
                                              last_unmap_address + 1 - cur_address) /
                                     PageSize;
 
-                                // HACK: Manually close the pages.
-                                HACK_ClosePages(cur_address, cur_pages);
-
                                 // Unmap.
                                 ASSERT(Operate(cur_address, cur_pages, KMemoryPermission::None,
                                                OperationType::Unmap)
@@ -1703,7 +1670,7 @@ Result KPageTable::MapPhysicalMemory(VAddr address, size_t size) {
                     // Release any remaining unmapped memory.
                     m_system.Kernel().MemoryManager().OpenFirst(pg_phys_addr, pg_pages);
                     m_system.Kernel().MemoryManager().Close(pg_phys_addr, pg_pages);
-                    for (++pg_it; pg_it != pg.Nodes().end(); ++pg_it) {
+                    for (++pg_it; pg_it != pg.end(); ++pg_it) {
                         m_system.Kernel().MemoryManager().OpenFirst(pg_it->GetAddress(),
                                                                     pg_it->GetNumPages());
                         m_system.Kernel().MemoryManager().Close(pg_it->GetAddress(),
@@ -1731,7 +1698,7 @@ Result KPageTable::MapPhysicalMemory(VAddr address, size_t size) {
                             // Check if we're at the end of the physical block.
                             if (pg_pages == 0) {
                                 // Ensure there are more pages to map.
-                                ASSERT(pg_it != pg.Nodes().end());
+                                ASSERT(pg_it != pg.end());
 
                                 // Advance our physical block.
                                 ++pg_it;
@@ -1742,10 +1709,7 @@ Result KPageTable::MapPhysicalMemory(VAddr address, size_t size) {
                             // Map whatever we can.
                             const size_t cur_pages = std::min(pg_pages, map_pages);
                             R_TRY(Operate(cur_address, cur_pages, KMemoryPermission::UserReadWrite,
-                                          OperationType::Map, pg_phys_addr));
-
-                            // HACK: Manually open the pages.
-                            HACK_OpenPages(pg_phys_addr, cur_pages);
+                                          OperationType::MapFirst, pg_phys_addr));
 
                             // Advance.
                             cur_address += cur_pages * PageSize;
@@ -1888,9 +1852,6 @@ Result KPageTable::UnmapPhysicalMemory(VAddr address, size_t size) {
                                               last_address + 1 - cur_address) /
                                      PageSize;
 
-            // HACK: Manually close the pages.
-            HACK_ClosePages(cur_address, cur_pages);
-
             // Unmap.
             ASSERT(Operate(cur_address, cur_pages, KMemoryPermission::None, OperationType::Unmap)
                        .IsSuccess());
@@ -1955,7 +1916,7 @@ Result KPageTable::MapMemory(VAddr dst_address, VAddr src_address, size_t size)
     R_TRY(dst_allocator_result);
 
     // Map the memory.
-    KPageGroup page_linked_list;
+    KPageGroup page_linked_list{m_kernel, m_block_info_manager};
     const size_t num_pages{size / PageSize};
     const KMemoryPermission new_src_perm = static_cast<KMemoryPermission>(
         KMemoryPermission::KernelRead | KMemoryPermission::NotMapped);
@@ -2022,14 +1983,14 @@ Result KPageTable::UnmapMemory(VAddr dst_address, VAddr src_address, size_t size
                                                      num_dst_allocator_blocks);
     R_TRY(dst_allocator_result);
 
-    KPageGroup src_pages;
-    KPageGroup dst_pages;
+    KPageGroup src_pages{m_kernel, m_block_info_manager};
+    KPageGroup dst_pages{m_kernel, m_block_info_manager};
     const size_t num_pages{size / PageSize};
 
     AddRegionToPages(src_address, num_pages, src_pages);
     AddRegionToPages(dst_address, num_pages, dst_pages);
 
-    R_UNLESS(dst_pages.IsEqual(src_pages), ResultInvalidMemoryRegion);
+    R_UNLESS(dst_pages.IsEquivalentTo(src_pages), ResultInvalidMemoryRegion);
 
     {
         auto block_guard = detail::ScopeExit([&] { MapPages(dst_address, dst_pages, dst_perm); });
@@ -2060,7 +2021,7 @@ Result KPageTable::MapPages(VAddr addr, const KPageGroup& page_linked_list,
 
     VAddr cur_addr{addr};
 
-    for (const auto& node : page_linked_list.Nodes()) {
+    for (const auto& node : page_linked_list) {
         if (const auto result{
                 Operate(cur_addr, node.GetNumPages(), perm, OperationType::Map, node.GetAddress())};
             result.IsError()) {
@@ -2160,7 +2121,7 @@ Result KPageTable::UnmapPages(VAddr addr, const KPageGroup& page_linked_list) {
 
     VAddr cur_addr{addr};
 
-    for (const auto& node : page_linked_list.Nodes()) {
+    for (const auto& node : page_linked_list) {
         if (const auto result{Operate(cur_addr, node.GetNumPages(), KMemoryPermission::None,
                                       OperationType::Unmap)};
             result.IsError()) {
@@ -2527,13 +2488,13 @@ Result KPageTable::SetHeapSize(VAddr* out, size_t size) {
     R_UNLESS(memory_reservation.Succeeded(), ResultLimitReached);
 
     // Allocate pages for the heap extension.
-    KPageGroup pg;
+    KPageGroup pg{m_kernel, m_block_info_manager};
     R_TRY(m_system.Kernel().MemoryManager().AllocateAndOpen(
         &pg, allocation_size / PageSize,
         KMemoryManager::EncodeOption(m_memory_pool, m_allocation_option)));
 
     // Clear all the newly allocated pages.
-    for (const auto& it : pg.Nodes()) {
+    for (const auto& it : pg) {
         std::memset(m_system.DeviceMemory().GetPointer<void>(it.GetAddress()), m_heap_fill_value,
                     it.GetSize());
     }
@@ -2610,11 +2571,23 @@ ResultVal<VAddr> KPageTable::AllocateAndMapMemory(size_t needed_num_pages, size_
     if (is_map_only) {
         R_TRY(Operate(addr, needed_num_pages, perm, OperationType::Map, map_addr));
     } else {
-        KPageGroup page_group;
-        R_TRY(m_system.Kernel().MemoryManager().AllocateForProcess(
-            &page_group, needed_num_pages,
-            KMemoryManager::EncodeOption(m_memory_pool, m_allocation_option), 0, 0));
-        R_TRY(Operate(addr, needed_num_pages, page_group, OperationType::MapGroup));
+        // Create a page group tohold the pages we allocate.
+        KPageGroup pg{m_kernel, m_block_info_manager};
+
+        R_TRY(m_system.Kernel().MemoryManager().AllocateAndOpen(
+            &pg, needed_num_pages,
+            KMemoryManager::EncodeOption(m_memory_pool, m_allocation_option)));
+
+        // Ensure that the page group is closed when we're done working with it.
+        SCOPE_EXIT({ pg.Close(); });
+
+        // Clear all pages.
+        for (const auto& it : pg) {
+            std::memset(m_system.DeviceMemory().GetPointer<void>(it.GetAddress()),
+                        m_heap_fill_value, it.GetSize());
+        }
+
+        R_TRY(Operate(addr, needed_num_pages, pg, OperationType::MapGroup));
     }
 
     // Update the blocks.
@@ -2795,19 +2768,28 @@ Result KPageTable::Operate(VAddr addr, size_t num_pages, const KPageGroup& page_
     ASSERT(num_pages > 0);
     ASSERT(num_pages == page_group.GetNumPages());
 
-    for (const auto& node : page_group.Nodes()) {
-        const size_t size{node.GetNumPages() * PageSize};
+    switch (operation) {
+    case OperationType::MapGroup: {
+        // We want to maintain a new reference to every page in the group.
+        KScopedPageGroup spg(page_group);
 
-        switch (operation) {
-        case OperationType::MapGroup:
+        for (const auto& node : page_group) {
+            const size_t size{node.GetNumPages() * PageSize};
+
+            // Map the pages.
             m_system.Memory().MapMemoryRegion(*m_page_table_impl, addr, size, node.GetAddress());
-            break;
-        default:
-            ASSERT(false);
-            break;
+
+            addr += size;
         }
 
-        addr += size;
+        // We succeeded! We want to persist the reference to the pages.
+        spg.CancelClose();
+
+        break;
+    }
+    default:
+        ASSERT(false);
+        break;
     }
 
     R_SUCCEED();
@@ -2822,13 +2804,29 @@ Result KPageTable::Operate(VAddr addr, size_t num_pages, KMemoryPermission perm,
     ASSERT(ContainsPages(addr, num_pages));
 
     switch (operation) {
-    case OperationType::Unmap:
+    case OperationType::Unmap: {
+        // Ensure that any pages we track close on exit.
+        KPageGroup pages_to_close{m_kernel, this->GetBlockInfoManager()};
+        SCOPE_EXIT({ pages_to_close.CloseAndReset(); });
+
+        this->AddRegionToPages(addr, num_pages, pages_to_close);
         m_system.Memory().UnmapRegion(*m_page_table_impl, addr, num_pages * PageSize);
         break;
+    }
+    case OperationType::MapFirst:
     case OperationType::Map: {
         ASSERT(map_addr);
         ASSERT(Common::IsAligned(map_addr, PageSize));
         m_system.Memory().MapMemoryRegion(*m_page_table_impl, addr, num_pages * PageSize, map_addr);
+
+        // Open references to pages, if we should.
+        if (IsHeapPhysicalAddress(m_kernel.MemoryLayout(), map_addr)) {
+            if (operation == OperationType::MapFirst) {
+                m_kernel.MemoryManager().OpenFirst(map_addr, num_pages);
+            } else {
+                m_kernel.MemoryManager().Open(map_addr, num_pages);
+            }
+        }
         break;
     }
     case OperationType::Separate: {

src/core/hle/kernel/k_page_table.h

@@ -107,6 +107,10 @@ public:
         return *m_page_table_impl;
     }
 
+    KBlockInfoManager* GetBlockInfoManager() {
+        return m_block_info_manager;
+    }
+
     bool CanContain(VAddr addr, size_t size, KMemoryState state) const;
 
 protected:
@@ -261,10 +265,6 @@ private:
     void CleanupForIpcClientOnServerSetupFailure(PageLinkedList* page_list, VAddr address,
                                                  size_t size, KMemoryPermission prot_perm);
 
-    // HACK: These will be removed once we automatically manage page reference counts.
-    void HACK_OpenPages(PAddr phys_addr, size_t num_pages);
-    void HACK_ClosePages(VAddr virt_addr, size_t num_pages);
-
     mutable KLightLock m_general_lock;
     mutable KLightLock m_map_physical_memory_lock;
 
@@ -488,6 +488,7 @@ private:
     std::unique_ptr<Common::PageTable> m_page_table_impl;
 
     Core::System& m_system;
+    KernelCore& m_kernel;
 };
 
 } // namespace Kernel

src/core/hle/kernel/k_shared_memory.cpp

@@ -13,10 +13,7 @@
 namespace Kernel {
 
 KSharedMemory::KSharedMemory(KernelCore& kernel_) : KAutoObjectWithSlabHeapAndContainer{kernel_} {}
-
-KSharedMemory::~KSharedMemory() {
-    kernel.GetSystemResourceLimit()->Release(LimitableResource::PhysicalMemoryMax, size);
-}
+KSharedMemory::~KSharedMemory() = default;
 
 Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* owner_process_,
                                  Svc::MemoryPermission owner_permission_,
@@ -49,7 +46,8 @@ Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* o
     R_UNLESS(physical_address != 0, ResultOutOfMemory);
 
     //! Insert the result into our page group.
-    page_group.emplace(physical_address, num_pages);
+    page_group.emplace(kernel, &kernel.GetSystemSystemResource().GetBlockInfoManager());
+    page_group->AddBlock(physical_address, num_pages);
 
     // Commit our reservation.
     memory_reservation.Commit();
@@ -62,7 +60,7 @@ Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* o
     is_initialized = true;
 
     // Clear all pages in the memory.
-    for (const auto& block : page_group->Nodes()) {
+    for (const auto& block : *page_group) {
         std::memset(device_memory_.GetPointer<void>(block.GetAddress()), 0, block.GetSize());
     }
 
@@ -71,13 +69,8 @@ Result KSharedMemory::Initialize(Core::DeviceMemory& device_memory_, KProcess* o
 
 void KSharedMemory::Finalize() {
     // Close and finalize the page group.
-    // page_group->Close();
-    // page_group->Finalize();
-
-    //! HACK: Manually close.
-    for (const auto& block : page_group->Nodes()) {
-        kernel.MemoryManager().Close(block.GetAddress(), block.GetNumPages());
-    }
+    page_group->Close();
+    page_group->Finalize();
 
     // Release the memory reservation.
     resource_limit->Release(LimitableResource::PhysicalMemoryMax, size);

src/core/hle/kernel/memory_types.h

@@ -14,4 +14,7 @@ constexpr std::size_t PageSize{1 << PageBits};
 
 using Page = std::array<u8, PageSize>;
 
+using KPhysicalAddress = PAddr;
+using KProcessAddress = VAddr;
+
 } // namespace Kernel

src/core/hle/kernel/svc.cpp

@@ -1485,7 +1485,7 @@ static Result MapProcessMemory(Core::System& system, VAddr dst_address, Handle p
              ResultInvalidMemoryRegion);
 
     // Create a new page group.
-    KPageGroup pg;
+    KPageGroup pg{system.Kernel(), dst_pt.GetBlockInfoManager()};
     R_TRY(src_pt.MakeAndOpenPageGroup(
         std::addressof(pg), src_address, size / PageSize, KMemoryState::FlagCanMapProcess,
         KMemoryState::FlagCanMapProcess, KMemoryPermission::None, KMemoryPermission::None,

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

@@ -312,8 +312,6 @@ void NVFlinger::Compose() {
 }
 
 s64 NVFlinger::GetNextTicks() const {
-    static constexpr s64 max_hertz = 120LL;
-
     const auto& settings = Settings::values;
     auto speed_scale = 1.f;
     if (settings.use_multi_core.GetValue()) {
@@ -327,9 +325,11 @@ s64 NVFlinger::GetNextTicks() const {
         }
     }
 
-    const auto next_ticks = ((1000000000 * (1LL << swap_interval)) / max_hertz);
+    // As an extension, treat nonpositive swap interval as framerate multiplier.
+    const f32 effective_fps = swap_interval <= 0 ? 120.f * static_cast<f32>(1 - swap_interval)
+                                                 : 60.f / static_cast<f32>(swap_interval);
 
-    return static_cast<s64>(speed_scale * static_cast<float>(next_ticks));
+    return static_cast<s64>(speed_scale * (1000000000.f / effective_fps));
 }
 
 } // namespace Service::NVFlinger

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

@@ -133,7 +133,7 @@ private:
     /// layers.
     u32 next_buffer_queue_id = 1;
 
-    u32 swap_interval = 1;
+    s32 swap_interval = 1;
 
     /// Event that handles screen composition.
     std::shared_ptr<Core::Timing::EventType> multi_composition_event;

src/core/internal_network/network.cpp

@@ -117,6 +117,8 @@ Errno TranslateNativeError(int e) {
         return Errno::NETUNREACH;
     case WSAEMSGSIZE:
         return Errno::MSGSIZE;
+    case WSAETIMEDOUT:
+        return Errno::TIMEDOUT;
     default:
         UNIMPLEMENTED_MSG("Unimplemented errno={}", e);
         return Errno::OTHER;
@@ -211,6 +213,8 @@ Errno TranslateNativeError(int e) {
         return Errno::NETUNREACH;
     case EMSGSIZE:
         return Errno::MSGSIZE;
+    case ETIMEDOUT:
+        return Errno::TIMEDOUT;
     default:
         UNIMPLEMENTED_MSG("Unimplemented errno={}", e);
         return Errno::OTHER;
@@ -226,7 +230,7 @@ Errno GetAndLogLastError() {
     int e = errno;
 #endif
     const Errno err = TranslateNativeError(e);
-    if (err == Errno::AGAIN) {
+    if (err == Errno::AGAIN || err == Errno::TIMEDOUT) {
         return err;
     }
     LOG_ERROR(Network, "Socket operation error: {}", Common::NativeErrorToString(e));

src/core/memory.cpp

@@ -436,7 +436,7 @@ struct Memory::Impl {
         }
 
         if (Settings::IsFastmemEnabled()) {
-            const bool is_read_enable = Settings::IsGPULevelHigh() || !cached;
+            const bool is_read_enable = !Settings::IsGPULevelExtreme() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
         }
 

src/input_common/drivers/tas_input.cpp

@@ -156,10 +156,12 @@ void Tas::RecordInput(u64 buttons, TasAnalog left_axis, TasAnalog right_axis) {
     };
 }
 
-std::tuple<TasState, size_t, size_t> Tas::GetStatus() const {
+std::tuple<TasState, size_t, std::array<size_t, PLAYER_NUMBER>> Tas::GetStatus() const {
     TasState state;
+    std::array<size_t, PLAYER_NUMBER> lengths{0};
     if (is_recording) {
-        return {TasState::Recording, 0, record_commands.size()};
+        lengths[0] = record_commands.size();
+        return {TasState::Recording, record_commands.size(), lengths};
     }
 
     if (is_running) {
@@ -168,7 +170,11 @@ std::tuple<TasState, size_t, size_t> Tas::GetStatus() const {
         state = TasState::Stopped;
     }
 
-    return {state, current_command, script_length};
+    for (size_t i = 0; i < PLAYER_NUMBER; i++) {
+        lengths[i] = commands[i].size();
+    }
+
+    return {state, current_command, lengths};
 }
 
 void Tas::UpdateThread() {

src/input_common/drivers/tas_input.h

@@ -124,7 +124,7 @@ public:
      * Current playback progress ;
      * Total length of script file currently loaded or being recorded
      */
-    std::tuple<TasState, size_t, size_t> GetStatus() const;
+    std::tuple<TasState, size_t, std::array<size_t, PLAYER_NUMBER>> GetStatus() const;
 
 private:
     enum class TasAxis : u8;

src/input_common/helpers/stick_from_buttons.cpp

@@ -13,11 +13,11 @@ class Stick final : public Common::Input::InputDevice {
 public:
     using Button = std::unique_ptr<Common::Input::InputDevice>;
 
-    Stick(Button up_, Button down_, Button left_, Button right_, Button modifier_,
+    Stick(Button up_, Button down_, Button left_, Button right_, Button modifier_, Button updater_,
           float modifier_scale_, float modifier_angle_)
         : up(std::move(up_)), down(std::move(down_)), left(std::move(left_)),
-          right(std::move(right_)), modifier(std::move(modifier_)), modifier_scale(modifier_scale_),
-          modifier_angle(modifier_angle_) {
+          right(std::move(right_)), modifier(std::move(modifier_)), updater(std::move(updater_)),
+          modifier_scale(modifier_scale_), modifier_angle(modifier_angle_) {
         up->SetCallback({
             .on_change =
                 [this](const Common::Input::CallbackStatus& callback_) {
@@ -48,6 +48,9 @@ public:
                     UpdateModButtonStatus(callback_);
                 },
         });
+        updater->SetCallback({
+            .on_change = [this](const Common::Input::CallbackStatus& callback_) { SoftUpdate(); },
+        });
         last_x_axis_value = 0.0f;
         last_y_axis_value = 0.0f;
     }
@@ -248,7 +251,7 @@ public:
         modifier->ForceUpdate();
     }
 
-    void SoftUpdate() override {
+    void SoftUpdate() {
         Common::Input::CallbackStatus status{
             .type = Common::Input::InputType::Stick,
             .stick_status = GetStatus(),
@@ -308,6 +311,7 @@ private:
     Button left;
     Button right;
     Button modifier;
+    Button updater;
     float modifier_scale{};
     float modifier_angle{};
     float angle{};
@@ -331,11 +335,12 @@ std::unique_ptr<Common::Input::InputDevice> StickFromButton::Create(
     auto left = Common::Input::CreateInputDeviceFromString(params.Get("left", null_engine));
     auto right = Common::Input::CreateInputDeviceFromString(params.Get("right", null_engine));
     auto modifier = Common::Input::CreateInputDeviceFromString(params.Get("modifier", null_engine));
+    auto updater = Common::Input::CreateInputDeviceFromString("engine:updater,button:0");
     auto modifier_scale = params.Get("modifier_scale", 0.5f);
     auto modifier_angle = params.Get("modifier_angle", 5.5f);
     return std::make_unique<Stick>(std::move(up), std::move(down), std::move(left),
-                                   std::move(right), std::move(modifier), modifier_scale,
-                                   modifier_angle);
+                                   std::move(right), std::move(modifier), std::move(updater),
+                                   modifier_scale, modifier_angle);
 }
 
 } // namespace InputCommon

src/input_common/input_mapping.cpp

@@ -76,7 +76,7 @@ void MappingFactory::RegisterButton(const MappingData& data) {
         break;
     case EngineInputType::Analog:
         // Ignore mouse axis when mapping buttons
-        if (data.engine == "mouse") {
+        if (data.engine == "mouse" && data.index != 4) {
             return;
         }
         new_input.Set("axis", data.index);

src/input_common/main.cpp

@@ -28,6 +28,28 @@
 
 namespace InputCommon {
 
+/// Dummy engine to get periodic updates
+class UpdateEngine final : public InputEngine {
+public:
+    explicit UpdateEngine(std::string input_engine_) : InputEngine(std::move(input_engine_)) {
+        PreSetController(identifier);
+    }
+
+    void PumpEvents() {
+        SetButton(identifier, 0, last_state);
+        last_state = !last_state;
+    }
+
+private:
+    static constexpr PadIdentifier identifier = {
+        .guid = Common::UUID{},
+        .port = 0,
+        .pad = 0,
+    };
+
+    bool last_state{};
+};
+
 struct InputSubsystem::Impl {
     template <typename Engine>
     void RegisterEngine(std::string name, std::shared_ptr<Engine>& engine) {
@@ -45,6 +67,7 @@ struct InputSubsystem::Impl {
     void Initialize() {
         mapping_factory = std::make_shared<MappingFactory>();
 
+        RegisterEngine("updater", update_engine);
         RegisterEngine("keyboard", keyboard);
         RegisterEngine("mouse", mouse);
         RegisterEngine("touch", touch_screen);
@@ -74,6 +97,7 @@ struct InputSubsystem::Impl {
     }
 
     void Shutdown() {
+        UnregisterEngine(update_engine);
         UnregisterEngine(keyboard);
         UnregisterEngine(mouse);
         UnregisterEngine(touch_screen);
@@ -252,6 +276,7 @@ struct InputSubsystem::Impl {
     }
 
     void PumpEvents() const {
+        update_engine->PumpEvents();
 #ifdef HAVE_SDL2
         sdl->PumpEvents();
 #endif
@@ -263,6 +288,7 @@ struct InputSubsystem::Impl {
 
     std::shared_ptr<MappingFactory> mapping_factory;
 
+    std::shared_ptr<UpdateEngine> update_engine;
     std::shared_ptr<Keyboard> keyboard;
     std::shared_ptr<Mouse> mouse;
     std::shared_ptr<TouchScreen> touch_screen;

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -321,8 +321,12 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
     case IR::Attribute::PositionY:
     case IR::Attribute::PositionZ:
     case IR::Attribute::PositionW:
-        return ctx.OpLoad(ctx.F32[1], AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position,
-                                                  ctx.Const(element)));
+        return ctx.OpLoad(
+            ctx.F32[1],
+            ctx.need_input_position_indirect
+                ? AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position, ctx.u32_zero_value,
+                              ctx.Const(element))
+                : AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position, ctx.Const(element)));
     case IR::Attribute::InstanceId:
         if (ctx.profile.support_vertex_instance_id) {
             return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.instance_id));

src/shader_recompiler/backend/spirv/emit_spirv_warp.cpp

@@ -58,11 +58,10 @@ Id SelectValue(EmitContext& ctx, Id in_range, Id value, Id src_thread_id) {
         ctx.OpGroupNonUniformShuffle(ctx.U32[1], SubgroupScope(ctx), value, src_thread_id), value);
 }
 
-Id GetUpperClamp(EmitContext& ctx, Id invocation_id, Id clamp) {
-    const Id thirty_two{ctx.Const(32u)};
-    const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, invocation_id, thirty_two)};
-    const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
-    return ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
+Id AddPartitionBase(EmitContext& ctx, Id thread_id) {
+    const Id partition_idx{ctx.OpShiftRightLogical(ctx.U32[1], GetThreadId(ctx), ctx.Const(5u))};
+    const Id partition_base{ctx.OpShiftLeftLogical(ctx.U32[1], partition_idx, ctx.Const(5u))};
+    return ctx.OpIAdd(ctx.U32[1], thread_id, partition_base);
 }
 } // Anonymous namespace
 
@@ -145,64 +144,63 @@ Id EmitSubgroupGeMask(EmitContext& ctx) {
 Id EmitShuffleIndex(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                     Id segmentation_mask) {
     const Id not_seg_mask{ctx.OpNot(ctx.U32[1], segmentation_mask)};
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        const Id thirty_two{ctx.Const(32u)};
-        const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, thread_id, thirty_two)};
-        const Id upper_index{ctx.OpIAdd(ctx.U32[1], thirty_two, index)};
-        const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
-        index = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_index, index);
-        clamp = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id min_thread_id{ComputeMinThreadId(ctx, thread_id, segmentation_mask)};
     const Id max_thread_id{ComputeMaxThreadId(ctx, min_thread_id, clamp, not_seg_mask)};
 
     const Id lhs{ctx.OpBitwiseAnd(ctx.U32[1], index, not_seg_mask)};
-    const Id src_thread_id{ctx.OpBitwiseOr(ctx.U32[1], lhs, min_thread_id)};
+    Id src_thread_id{ctx.OpBitwiseOr(ctx.U32[1], lhs, min_thread_id)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleUp(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                  Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSGreaterThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleDown(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                    Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleButterfly(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                         Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -544,7 +544,7 @@ void EmitContext::DefineCommonTypes(const Info& info) {
         U16 = Name(TypeInt(16, false), "u16");
         S16 = Name(TypeInt(16, true), "s16");
     }
-    if (info.uses_int64) {
+    if (info.uses_int64 && profile.support_int64) {
         AddCapability(spv::Capability::Int64);
         U64 = Name(TypeInt(64, false), "u64");
     }
@@ -721,9 +721,21 @@ void EmitContext::DefineAttributeMemAccess(const Info& info) {
         size_t label_index{0};
         if (info.loads.AnyComponent(IR::Attribute::PositionX)) {
             AddLabel(labels[label_index]);
-            const Id pointer{is_array
-                                 ? OpAccessChain(input_f32, input_position, vertex, masked_index)
-                                 : OpAccessChain(input_f32, input_position, masked_index)};
+            const Id pointer{[&]() {
+                if (need_input_position_indirect) {
+                    if (is_array)
+                        return OpAccessChain(input_f32, input_position, vertex, u32_zero_value,
+                                             masked_index);
+                    else
+                        return OpAccessChain(input_f32, input_position, u32_zero_value,
+                                             masked_index);
+                } else {
+                    if (is_array)
+                        return OpAccessChain(input_f32, input_position, vertex, masked_index);
+                    else
+                        return OpAccessChain(input_f32, input_position, masked_index);
+                }
+            }()};
             const Id result{OpLoad(F32[1], pointer)};
             OpReturnValue(result);
             ++label_index;
@@ -1367,12 +1379,25 @@ void EmitContext::DefineInputs(const IR::Program& program) {
         Decorate(layer, spv::Decoration::Flat);
     }
     if (loads.AnyComponent(IR::Attribute::PositionX)) {
-        const bool is_fragment{stage != Stage::Fragment};
-        const spv::BuiltIn built_in{is_fragment ? spv::BuiltIn::Position : spv::BuiltIn::FragCoord};
-        input_position = DefineInput(*this, F32[4], true, built_in);
-        if (profile.support_geometry_shader_passthrough) {
-            if (info.passthrough.AnyComponent(IR::Attribute::PositionX)) {
-                Decorate(input_position, spv::Decoration::PassthroughNV);
+        const bool is_fragment{stage == Stage::Fragment};
+        if (!is_fragment && profile.has_broken_spirv_position_input) {
+            need_input_position_indirect = true;
+
+            const Id input_position_struct = TypeStruct(F32[4]);
+            input_position = DefineInput(*this, input_position_struct, true);
+
+            MemberDecorate(input_position_struct, 0, spv::Decoration::BuiltIn,
+                           static_cast<unsigned>(spv::BuiltIn::Position));
+            Decorate(input_position_struct, spv::Decoration::Block);
+        } else {
+            const spv::BuiltIn built_in{is_fragment ? spv::BuiltIn::FragCoord
+                                                    : spv::BuiltIn::Position};
+            input_position = DefineInput(*this, F32[4], true, built_in);
+
+            if (profile.support_geometry_shader_passthrough) {
+                if (info.passthrough.AnyComponent(IR::Attribute::PositionX)) {
+                    Decorate(input_position, spv::Decoration::PassthroughNV);
+                }
             }
         }
     }

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -280,6 +280,7 @@ public:
     Id write_global_func_u32x2{};
     Id write_global_func_u32x4{};
 
+    bool need_input_position_indirect{};
     Id input_position{};
     std::array<Id, 32> input_generics{};
 

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -171,6 +171,70 @@ std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
     }
     return mapping;
 }
+
+void EmitGeometryPassthrough(IR::IREmitter& ir, const IR::Program& program,
+                             const Shader::VaryingState& passthrough_mask,
+                             bool passthrough_position,
+                             std::optional<IR::Attribute> passthrough_layer_attr) {
+    for (u32 i = 0; i < program.output_vertices; i++) {
+        // Assign generics from input
+        for (u32 j = 0; j < 32; j++) {
+            if (!passthrough_mask.Generic(j)) {
+                continue;
+            }
+
+            const IR::Attribute attr = IR::Attribute::Generic0X + (j * 4);
+            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+        }
+
+        if (passthrough_position) {
+            // Assign position from input
+            const IR::Attribute attr = IR::Attribute::PositionX;
+            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+        }
+
+        if (passthrough_layer_attr) {
+            // Assign layer
+            ir.SetAttribute(IR::Attribute::Layer, ir.GetAttribute(*passthrough_layer_attr),
+                            ir.Imm32(0));
+        }
+
+        // Emit vertex
+        ir.EmitVertex(ir.Imm32(0));
+    }
+    ir.EndPrimitive(ir.Imm32(0));
+}
+
+u32 GetOutputTopologyVertices(OutputTopology output_topology) {
+    switch (output_topology) {
+    case OutputTopology::PointList:
+        return 1;
+    case OutputTopology::LineStrip:
+        return 2;
+    default:
+        return 3;
+    }
+}
+
+void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInfo& host_info) {
+    for (IR::Block* const block : program.blocks) {
+        for (IR::Inst& inst : block->Instructions()) {
+            if (inst.GetOpcode() == IR::Opcode::Epilogue) {
+                IR::IREmitter ir{*block, IR::Block::InstructionList::s_iterator_to(inst)};
+                EmitGeometryPassthrough(
+                    ir, program, program.info.passthrough,
+                    program.info.passthrough.AnyComponent(IR::Attribute::PositionX), {});
+            }
+        }
+    }
+}
+
 } // Anonymous namespace
 
 IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
@@ -195,9 +259,14 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
         program.is_geometry_passthrough = sph.common0.geometry_passthrough != 0;
         if (program.is_geometry_passthrough) {
             const auto& mask{env.GpPassthroughMask()};
-            for (size_t i = 0; i < program.info.passthrough.mask.size(); ++i) {
+            for (size_t i = 0; i < mask.size() * 32; ++i) {
                 program.info.passthrough.mask[i] = ((mask[i / 32] >> (i % 32)) & 1) == 0;
             }
+
+            if (!host_info.support_geometry_shader_passthrough) {
+                program.output_vertices = GetOutputTopologyVertices(program.output_topology);
+                LowerGeometryPassthrough(program, host_info);
+            }
         }
         break;
     }
@@ -223,7 +292,7 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
 
     Optimization::PositionPass(env, program);
 
-    Optimization::GlobalMemoryToStorageBufferPass(program);
+    Optimization::GlobalMemoryToStorageBufferPass(program, host_info);
     Optimization::TexturePass(env, program, host_info);
 
     if (Settings::values.resolution_info.active) {
@@ -342,17 +411,7 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
     IR::Program program;
     program.stage = Stage::Geometry;
     program.output_topology = output_topology;
-    switch (output_topology) {
-    case OutputTopology::PointList:
-        program.output_vertices = 1;
-        break;
-    case OutputTopology::LineStrip:
-        program.output_vertices = 2;
-        break;
-    default:
-        program.output_vertices = 3;
-        break;
-    }
+    program.output_vertices = GetOutputTopologyVertices(output_topology);
 
     program.is_geometry_passthrough = false;
     program.info.loads.mask = source_program.info.stores.mask;
@@ -366,35 +425,8 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
     node.data.block = current_block;
 
     IR::IREmitter ir{*current_block};
-    for (u32 i = 0; i < program.output_vertices; i++) {
-        // Assign generics from input
-        for (u32 j = 0; j < 32; j++) {
-            if (!program.info.stores.Generic(j)) {
-                continue;
-            }
-
-            const IR::Attribute attr = IR::Attribute::Generic0X + (j * 4);
-            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
-        }
-
-        // Assign position from input
-        const IR::Attribute attr = IR::Attribute::PositionX;
-        ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
-
-        // Assign layer
-        ir.SetAttribute(IR::Attribute::Layer, ir.GetAttribute(source_program.info.emulated_layer),
-                        ir.Imm32(0));
-
-        // Emit vertex
-        ir.EmitVertex(ir.Imm32(0));
-    }
-    ir.EndPrimitive(ir.Imm32(0));
+    EmitGeometryPassthrough(ir, program, program.info.stores, true,
+                            source_program.info.emulated_layer);
 
     IR::Block* return_block{block_pool.Create(inst_pool)};
     IR::IREmitter{*return_block}.Epilogue();

src/shader_recompiler/host_translate_info.h

@@ -15,6 +15,9 @@ struct HostTranslateInfo {
     bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
     bool support_snorm_render_buffer{};  ///< True when the device supports SNORM render buffers
     bool support_viewport_index_layer{}; ///< True when the device supports gl_Layer in VS
+    u32 min_ssbo_alignment{};            ///< Minimum alignment supported by the device for SSBOs
+    bool support_geometry_shader_passthrough{}; ///< True when the device supports geometry
+                                                ///< passthrough shaders
 };
 
 } // namespace Shader

src/shader_recompiler/ir_opt/global_memory_to_storage_buffer_pass.cpp

@@ -11,6 +11,7 @@
 #include "shader_recompiler/frontend/ir/breadth_first_search.h"
 #include "shader_recompiler/frontend/ir/ir_emitter.h"
 #include "shader_recompiler/frontend/ir/value.h"
+#include "shader_recompiler/host_translate_info.h"
 #include "shader_recompiler/ir_opt/passes.h"
 
 namespace Shader::Optimization {
@@ -402,7 +403,7 @@ void CollectStorageBuffers(IR::Block& block, IR::Inst& inst, StorageInfo& info)
 }
 
 /// Returns the offset in indices (not bytes) for an equivalent storage instruction
-IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer) {
+IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer, u32 alignment) {
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::U32 offset;
     if (const std::optional<LowAddrInfo> low_addr{TrackLowAddress(&inst)}) {
@@ -415,7 +416,10 @@ IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer
     }
     // Subtract the least significant 32 bits from the guest offset. The result is the storage
     // buffer offset in bytes.
-    const IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
+    IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
+
+    // Align the offset base to match the host alignment requirements
+    low_cbuf = ir.BitwiseAnd(low_cbuf, ir.Imm32(~(alignment - 1U)));
     return ir.ISub(offset, low_cbuf);
 }
 
@@ -510,7 +514,7 @@ void Replace(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
 }
 } // Anonymous namespace
 
-void GlobalMemoryToStorageBufferPass(IR::Program& program) {
+void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info) {
     StorageInfo info;
     for (IR::Block* const block : program.post_order_blocks) {
         for (IR::Inst& inst : block->Instructions()) {
@@ -534,7 +538,8 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program) {
         const IR::U32 index{IR::Value{static_cast<u32>(info.set.index_of(it))}};
         IR::Block* const block{storage_inst.block};
         IR::Inst* const inst{storage_inst.inst};
-        const IR::U32 offset{StorageOffset(*block, *inst, storage_buffer)};
+        const IR::U32 offset{
+            StorageOffset(*block, *inst, storage_buffer, host_info.min_ssbo_alignment)};
         Replace(*block, *inst, index, offset);
     }
 }

src/shader_recompiler/ir_opt/passes.h

@@ -15,7 +15,7 @@ namespace Shader::Optimization {
 void CollectShaderInfoPass(Environment& env, IR::Program& program);
 void ConstantPropagationPass(Environment& env, IR::Program& program);
 void DeadCodeEliminationPass(IR::Program& program);
-void GlobalMemoryToStorageBufferPass(IR::Program& program);
+void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info);
 void IdentityRemovalPass(IR::Program& program);
 void LowerFp16ToFp32(IR::Program& program);
 void LowerInt64ToInt32(IR::Program& program);

src/shader_recompiler/profile.h

@@ -55,6 +55,8 @@ struct Profile {
 
     /// OpFClamp is broken and OpFMax + OpFMin should be used instead
     bool has_broken_spirv_clamp{};
+    /// The Position builtin needs to be wrapped in a struct when used as an input
+    bool has_broken_spirv_position_input{};
     /// Offset image operands with an unsigned type do not work
     bool has_broken_unsigned_image_offsets{};
     /// Signed instructions with unsigned data types are misinterpreted

src/shader_recompiler/shader_info.h

@@ -65,6 +65,8 @@ enum class Interpolation {
 struct ConstantBufferDescriptor {
     u32 index;
     u32 count;
+
+    auto operator<=>(const ConstantBufferDescriptor&) const = default;
 };
 
 struct StorageBufferDescriptor {
@@ -72,6 +74,8 @@ struct StorageBufferDescriptor {
     u32 cbuf_offset;
     u32 count;
     bool is_written;
+
+    auto operator<=>(const StorageBufferDescriptor&) const = default;
 };
 
 struct TextureBufferDescriptor {
@@ -84,6 +88,8 @@ struct TextureBufferDescriptor {
     u32 secondary_shift_left;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const TextureBufferDescriptor&) const = default;
 };
 using TextureBufferDescriptors = boost::container::small_vector<TextureBufferDescriptor, 6>;
 
@@ -95,6 +101,8 @@ struct ImageBufferDescriptor {
     u32 cbuf_offset;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const ImageBufferDescriptor&) const = default;
 };
 using ImageBufferDescriptors = boost::container::small_vector<ImageBufferDescriptor, 2>;
 
@@ -110,6 +118,8 @@ struct TextureDescriptor {
     u32 secondary_shift_left;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const TextureDescriptor&) const = default;
 };
 using TextureDescriptors = boost::container::small_vector<TextureDescriptor, 12>;
 
@@ -122,6 +132,8 @@ struct ImageDescriptor {
     u32 cbuf_offset;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const ImageDescriptor&) const = default;
 };
 using ImageDescriptors = boost::container::small_vector<ImageDescriptor, 4>;
 

src/tests/video_core/buffer_base.cpp

@@ -538,7 +538,7 @@ TEST_CASE("BufferBase: Cached write downloads") {
     int num = 0;
     buffer.ForEachDownloadRangeAndClear(c, WORD, [&](u64 offset, u64 size) { ++num; });
     buffer.ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
-    REQUIRE(num == 0);
+    REQUIRE(num == 1);
     REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
     REQUIRE(!buffer.IsRegionGpuModified(c + PAGE, PAGE));
     buffer.FlushCachedWrites();

src/video_core/CMakeLists.txt

@@ -85,6 +85,7 @@ add_library(video_core STATIC
     gpu.h
     gpu_thread.cpp
     gpu_thread.h
+    invalidation_accumulator.h
     memory_manager.cpp
     memory_manager.h
     precompiled_headers.h
@@ -190,6 +191,8 @@ add_library(video_core STATIC
     renderer_vulkan/vk_texture_cache.cpp
     renderer_vulkan/vk_texture_cache.h
     renderer_vulkan/vk_texture_cache_base.cpp
+    renderer_vulkan/vk_turbo_mode.cpp
+    renderer_vulkan/vk_turbo_mode.h
     renderer_vulkan/vk_update_descriptor.cpp
     renderer_vulkan/vk_update_descriptor.h
     shader_cache.cpp

src/video_core/buffer_cache/buffer_base.h

@@ -430,7 +430,7 @@ private:
         if (query_begin >= SizeBytes() || size < 0) {
             return;
         }
-        u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] u64* const untracked_words = Array<Type::Untracked>();
         u64* const state_words = Array<type>();
         const u64 query_end = query_begin + std::min(static_cast<u64>(size), SizeBytes());
         u64* const words_begin = state_words + query_begin / BYTES_PER_WORD;
@@ -483,7 +483,7 @@ private:
                 NotifyRasterizer<true>(word_index, current_bits, ~u64{0});
             }
             // Exclude CPU modified pages when visiting GPU pages
-            const u64 word = current_word & ~(type == Type::GPU ? untracked_words[word_index] : 0);
+            const u64 word = current_word;
             u64 page = page_begin;
             page_begin = 0;
 
@@ -531,7 +531,7 @@ private:
     [[nodiscard]] bool IsRegionModified(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        const u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -539,8 +539,7 @@ private:
         const u64 page_limit = Common::DivCeil(offset + size, BYTES_PER_PAGE);
         u64 page_index = (offset / BYTES_PER_PAGE) % PAGES_PER_WORD;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index, page_index = 0) {
-            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
-            const u64 word = state_words[word_index] & ~off_word;
+            const u64 word = state_words[word_index];
             if (word == 0) {
                 continue;
             }
@@ -564,7 +563,7 @@ private:
     [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        const u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -574,8 +573,7 @@ private:
         u64 begin = std::numeric_limits<u64>::max();
         u64 end = 0;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index) {
-            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
-            const u64 word = state_words[word_index] & ~off_word;
+            const u64 word = state_words[word_index];
             if (word == 0) {
                 continue;
             }

src/video_core/buffer_cache/buffer_cache.h

@@ -1938,14 +1938,21 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
                                                                       bool is_written) const {
     const GPUVAddr gpu_addr = gpu_memory->Read<u64>(ssbo_addr);
     const u32 size = gpu_memory->Read<u32>(ssbo_addr + 8);
-    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+    const u32 alignment = runtime.GetStorageBufferAlignment();
+
+    const GPUVAddr aligned_gpu_addr = Common::AlignDown(gpu_addr, alignment);
+    const u32 aligned_size =
+        Common::AlignUp(static_cast<u32>(gpu_addr - aligned_gpu_addr) + size, alignment);
+
+    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(aligned_gpu_addr);
     if (!cpu_addr || size == 0) {
         return NULL_BINDING;
     }
-    const VAddr cpu_end = Common::AlignUp(*cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
+
+    const VAddr cpu_end = Common::AlignUp(*cpu_addr + aligned_size, Core::Memory::YUZU_PAGESIZE);
     const Binding binding{
         .cpu_addr = *cpu_addr,
-        .size = is_written ? size : static_cast<u32>(cpu_end - *cpu_addr),
+        .size = is_written ? aligned_size : static_cast<u32>(cpu_end - *cpu_addr),
         .buffer_id = BufferId{},
     };
     return binding;

src/video_core/engines/engine_upload.cpp

@@ -76,7 +76,7 @@ void State::ProcessData(std::span<const u8> read_buffer) {
                                        regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,
                                        x_elements, regs.line_count, regs.dest.BlockHeight(),
                                        regs.dest.BlockDepth(), regs.line_length_in);
-        memory_manager.WriteBlock(address, tmp_buffer.data(), dst_size);
+        memory_manager.WriteBlockCached(address, tmp_buffer.data(), dst_size);
     }
 }
 

src/video_core/engines/fermi_2d.cpp

@@ -6,6 +6,7 @@
 #include "common/microprofile.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/engines/sw_blitter/blitter.h"
+#include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/surface.h"
 #include "video_core/textures/decoders.h"
@@ -20,8 +21,8 @@ namespace Tegra::Engines {
 
 using namespace Texture;
 
-Fermi2D::Fermi2D(MemoryManager& memory_manager_) {
-    sw_blitter = std::make_unique<Blitter::SoftwareBlitEngine>(memory_manager_);
+Fermi2D::Fermi2D(MemoryManager& memory_manager_) : memory_manager{memory_manager_} {
+    sw_blitter = std::make_unique<Blitter::SoftwareBlitEngine>(memory_manager);
     // Nvidia's OpenGL driver seems to assume these values
     regs.src.depth = 1;
     regs.dst.depth = 1;
@@ -104,6 +105,7 @@ void Fermi2D::Blit() {
         config.src_x0 = 0;
     }
 
+    memory_manager.FlushCaching();
     if (!rasterizer->AccelerateSurfaceCopy(src, regs.dst, config)) {
         sw_blitter->Blit(src, regs.dst, config);
     }

src/video_core/engines/fermi_2d.h

@@ -305,6 +305,7 @@ public:
 private:
     VideoCore::RasterizerInterface* rasterizer = nullptr;
     std::unique_ptr<Blitter::SoftwareBlitEngine> sw_blitter;
+    MemoryManager& memory_manager;
 
     /// Performs the copy from the source surface to the destination surface as configured in the
     /// registers.

src/video_core/engines/maxwell_3d.cpp

@@ -467,7 +467,7 @@ void Maxwell3D::ProcessMacroBind(u32 data) {
 }
 
 void Maxwell3D::ProcessFirmwareCall4() {
-    LOG_WARNING(HW_GPU, "(STUBBED) called");
+    LOG_DEBUG(HW_GPU, "(STUBBED) called");
 
     // Firmware call 4 is a blob that changes some registers depending on its parameters.
     // These registers don't affect emulation and so are stubbed by setting 0xd00 to 1.
@@ -485,11 +485,6 @@ void Maxwell3D::StampQueryResult(u64 payload, bool long_query) {
 }
 
 void Maxwell3D::ProcessQueryGet() {
-    // TODO(Subv): Support the other query units.
-    if (regs.report_semaphore.query.location != Regs::ReportSemaphore::Location::All) {
-        LOG_DEBUG(HW_GPU, "Locations other than ALL are unimplemented");
-    }
-
     switch (regs.report_semaphore.query.operation) {
     case Regs::ReportSemaphore::Operation::Release:
         if (regs.report_semaphore.query.short_query != 0) {
@@ -649,7 +644,7 @@ void Maxwell3D::ProcessCBMultiData(const u32* start_base, u32 amount) {
 
     const GPUVAddr address{buffer_address + regs.const_buffer.offset};
     const size_t copy_size = amount * sizeof(u32);
-    memory_manager.WriteBlock(address, start_base, copy_size);
+    memory_manager.WriteBlockCached(address, start_base, copy_size);
 
     // Increment the current buffer position.
     regs.const_buffer.offset += static_cast<u32>(copy_size);

src/video_core/engines/maxwell_dma.cpp

@@ -69,7 +69,7 @@ void MaxwellDMA::Launch() {
     if (launch.multi_line_enable) {
         const bool is_src_pitch = launch.src_memory_layout == LaunchDMA::MemoryLayout::PITCH;
         const bool is_dst_pitch = launch.dst_memory_layout == LaunchDMA::MemoryLayout::PITCH;
-
+        memory_manager.FlushCaching();
         if (!is_src_pitch && !is_dst_pitch) {
             // If both the source and the destination are in block layout, assert.
             CopyBlockLinearToBlockLinear();
@@ -104,6 +104,7 @@ void MaxwellDMA::Launch() {
                                             reinterpret_cast<u8*>(tmp_buffer.data()),
                                             regs.line_length_in * sizeof(u32));
         } else {
+            memory_manager.FlushCaching();
             const auto convert_linear_2_blocklinear_addr = [](u64 address) {
                 return (address & ~0x1f0ULL) | ((address & 0x40) >> 2) | ((address & 0x10) << 1) |
                        ((address & 0x180) >> 1) | ((address & 0x20) << 3);
@@ -121,8 +122,8 @@ void MaxwellDMA::Launch() {
                     memory_manager.ReadBlockUnsafe(
                         convert_linear_2_blocklinear_addr(regs.offset_in + offset),
                         tmp_buffer.data(), tmp_buffer.size());
-                    memory_manager.WriteBlock(regs.offset_out + offset, tmp_buffer.data(),
-                                              tmp_buffer.size());
+                    memory_manager.WriteBlockCached(regs.offset_out + offset, tmp_buffer.data(),
+                                                    tmp_buffer.size());
                 }
             } else if (is_src_pitch && !is_dst_pitch) {
                 UNIMPLEMENTED_IF(regs.line_length_in % 16 != 0);
@@ -132,7 +133,7 @@ void MaxwellDMA::Launch() {
                 for (u32 offset = 0; offset < regs.line_length_in; offset += 16) {
                     memory_manager.ReadBlockUnsafe(regs.offset_in + offset, tmp_buffer.data(),
                                                    tmp_buffer.size());
-                    memory_manager.WriteBlock(
+                    memory_manager.WriteBlockCached(
                         convert_linear_2_blocklinear_addr(regs.offset_out + offset),
                         tmp_buffer.data(), tmp_buffer.size());
                 }
@@ -141,8 +142,8 @@ void MaxwellDMA::Launch() {
                     std::vector<u8> tmp_buffer(regs.line_length_in);
                     memory_manager.ReadBlockUnsafe(regs.offset_in, tmp_buffer.data(),
                                                    regs.line_length_in);
-                    memory_manager.WriteBlock(regs.offset_out, tmp_buffer.data(),
-                                              regs.line_length_in);
+                    memory_manager.WriteBlockCached(regs.offset_out, tmp_buffer.data(),
+                                                    regs.line_length_in);
                 }
             }
         }
@@ -204,7 +205,7 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
                      src_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
                      regs.pitch_out);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::CopyPitchToBlockLinear() {
@@ -256,7 +257,7 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
                    dst_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
                    regs.pitch_in);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::FastCopyBlockLinearToPitch() {
@@ -287,7 +288,7 @@ void MaxwellDMA::FastCopyBlockLinearToPitch() {
                      regs.src_params.block_size.height, regs.src_params.block_size.depth,
                      regs.pitch_out);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::CopyBlockLinearToBlockLinear() {
@@ -347,7 +348,7 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
                    dst.depth, dst_x_offset, dst.origin.y, x_elements, regs.line_count,
                    dst.block_size.height, dst.block_size.depth, pitch);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::ReleaseSemaphore() {

src/video_core/host_shaders/CMakeLists.txt

@@ -47,6 +47,7 @@ set(SHADER_FILES
     vulkan_present_scaleforce_fp16.frag
     vulkan_present_scaleforce_fp32.frag
     vulkan_quad_indexed.comp
+    vulkan_turbo_mode.comp
     vulkan_uint8.comp
 )
 

src/video_core/host_shaders/vulkan_turbo_mode.comp

@@ -0,0 +1,29 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#version 460 core
+
+layout (local_size_x = 16, local_size_y = 8, local_size_z = 1) in;
+
+layout (binding = 0) buffer ThreadData {
+    uint data[];
+};
+
+uint xorshift32(uint x) {
+    x ^= x << 13;
+    x ^= x >> 17;
+    x ^= x << 5;
+    return x;
+}
+
+uint getGlobalIndex() {
+    return gl_GlobalInvocationID.x + gl_GlobalInvocationID.y * gl_WorkGroupSize.y * gl_NumWorkGroups.y;
+}
+
+void main() {
+    uint myIndex = xorshift32(getGlobalIndex());
+    uint otherIndex = xorshift32(myIndex);
+
+    uint otherValue = atomicAdd(data[otherIndex % data.length()], 0) + 1;
+    atomicAdd(data[myIndex % data.length()], otherValue);
+}

src/video_core/invalidation_accumulator.h

@@ -0,0 +1,79 @@
+// SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+
+namespace VideoCommon {
+
+class InvalidationAccumulator {
+public:
+    InvalidationAccumulator() = default;
+    ~InvalidationAccumulator() = default;
+
+    void Add(GPUVAddr address, size_t size) {
+        const auto reset_values = [&]() {
+            if (has_collected) {
+                buffer.emplace_back(start_address, accumulated_size);
+            }
+            start_address = address;
+            accumulated_size = size;
+            last_collection = start_address + size;
+        };
+        if (address >= start_address && address + size <= last_collection) [[likely]] {
+            return;
+        }
+        size = ((address + size + atomicity_size_mask) & atomicity_mask) - address;
+        address = address & atomicity_mask;
+        if (!has_collected) [[unlikely]] {
+            reset_values();
+            has_collected = true;
+            return;
+        }
+        if (address != last_collection) [[unlikely]] {
+            reset_values();
+            return;
+        }
+        accumulated_size += size;
+        last_collection += size;
+    }
+
+    void Clear() {
+        buffer.clear();
+        start_address = 0;
+        last_collection = 0;
+        has_collected = false;
+    }
+
+    bool AnyAccumulated() const {
+        return has_collected;
+    }
+
+    template <typename Func>
+    void Callback(Func&& func) {
+        if (!has_collected) {
+            return;
+        }
+        buffer.emplace_back(start_address, accumulated_size);
+        for (auto& [address, size] : buffer) {
+            func(address, size);
+        }
+    }
+
+private:
+    static constexpr size_t atomicity_bits = 5;
+    static constexpr size_t atomicity_size = 1ULL << atomicity_bits;
+    static constexpr size_t atomicity_size_mask = atomicity_size - 1;
+    static constexpr size_t atomicity_mask = ~atomicity_size_mask;
+    GPUVAddr start_address{};
+    GPUVAddr last_collection{};
+    size_t accumulated_size{};
+    bool has_collected{};
+    std::vector<std::pair<VAddr, size_t>> buffer;
+};
+
+} // namespace VideoCommon

src/video_core/macro/macro_hle.cpp

@@ -50,38 +50,6 @@ protected:
     Maxwell3D& maxwell3d;
 };
 
-class HLE_DrawArrays final : public HLEMacroImpl {
-public:
-    explicit HLE_DrawArrays(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
-
-    void Execute(const std::vector<u32>& parameters, [[maybe_unused]] u32 method) override {
-        maxwell3d.RefreshParameters();
-
-        auto topology = static_cast<Maxwell3D::Regs::PrimitiveTopology>(parameters[0]);
-        maxwell3d.draw_manager->DrawArray(topology, parameters[1], parameters[2],
-                                          maxwell3d.regs.global_base_instance_index, 1);
-    }
-};
-
-class HLE_DrawIndexed final : public HLEMacroImpl {
-public:
-    explicit HLE_DrawIndexed(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
-
-    void Execute(const std::vector<u32>& parameters, [[maybe_unused]] u32 method) override {
-        maxwell3d.RefreshParameters();
-        maxwell3d.regs.index_buffer.start_addr_high = parameters[1];
-        maxwell3d.regs.index_buffer.start_addr_low = parameters[2];
-        maxwell3d.regs.index_buffer.format =
-            static_cast<Engines::Maxwell3D::Regs::IndexFormat>(parameters[3]);
-        maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-
-        auto topology = static_cast<Maxwell3D::Regs::PrimitiveTopology>(parameters[0]);
-        maxwell3d.draw_manager->DrawIndex(topology, 0, parameters[4],
-                                          maxwell3d.regs.global_base_vertex_index,
-                                          maxwell3d.regs.global_base_instance_index, 1);
-    }
-};
-
 /*
  * @note: these macros have two versions, a normal and extended version, with the extended version
  * also assigning the base vertex/instance.
@@ -497,11 +465,6 @@ public:
 } // Anonymous namespace
 
 HLEMacro::HLEMacro(Maxwell3D& maxwell3d_) : maxwell3d{maxwell3d_} {
-    builders.emplace(0xDD6A7FA92A7D2674ULL,
-                     std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
-                         [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
-                             return std::make_unique<HLE_DrawArrays>(maxwell3d__);
-                         }));
     builders.emplace(0x0D61FC9FAAC9FCADULL,
                      std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
                          [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
@@ -512,11 +475,6 @@ HLEMacro::HLEMacro(Maxwell3D& maxwell3d_) : maxwell3d{maxwell3d_} {
                          [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
                              return std::make_unique<HLE_DrawArraysIndirect<true>>(maxwell3d__);
                          }));
-    builders.emplace(0x2DB33AADB741839CULL,
-                     std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
-                         [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {
-                             return std::make_unique<HLE_DrawIndexed>(maxwell3d__);
-                         }));
     builders.emplace(0x771BB18C62444DA0ULL,
                      std::function<std::unique_ptr<CachedMacro>(Maxwell3D&)>(
                          [](Maxwell3D& maxwell3d__) -> std::unique_ptr<CachedMacro> {

src/video_core/memory_manager.cpp

@@ -6,11 +6,13 @@
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
+#include "common/settings.h"
 #include "core/core.h"
 #include "core/device_memory.h"
 #include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_process.h"
 #include "core/memory.h"
+#include "video_core/invalidation_accumulator.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
@@ -26,7 +28,8 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64
       entries{}, big_entries{}, page_table{address_space_bits, address_space_bits + page_bits - 38,
                                            page_bits != big_page_bits ? page_bits : 0},
       kind_map{PTEKind::INVALID}, unique_identifier{unique_identifier_generator.fetch_add(
-                                      1, std::memory_order_acq_rel)} {
+                                      1, std::memory_order_acq_rel)},
+      accumulator{std::make_unique<VideoCommon::InvalidationAccumulator>()} {
     address_space_size = 1ULL << address_space_bits;
     page_size = 1ULL << page_bits;
     page_mask = page_size - 1ULL;
@@ -43,6 +46,11 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64
     big_page_table_cpu.resize(big_page_table_size);
     big_page_continous.resize(big_page_table_size / continous_bits, 0);
     entries.resize(page_table_size / 32, 0);
+    if (!Settings::IsGPULevelExtreme() && Settings::IsFastmemEnabled()) {
+        fastmem_arena = system.DeviceMemory().buffer.VirtualBasePointer();
+    } else {
+        fastmem_arena = nullptr;
+    }
 }
 
 MemoryManager::~MemoryManager() = default;
@@ -185,15 +193,12 @@ void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
     if (size == 0) {
         return;
     }
-    const auto submapped_ranges = GetSubmappedRange(gpu_addr, size);
+    GetSubmappedRangeImpl<false>(gpu_addr, size, page_stash);
 
-    for (const auto& [map_addr, map_size] : submapped_ranges) {
-        // Flush and invalidate through the GPU interface, to be asynchronous if possible.
-        const std::optional<VAddr> cpu_addr = GpuToCpuAddress(map_addr);
-        ASSERT(cpu_addr);
-
-        rasterizer->UnmapMemory(*cpu_addr, map_size);
+    for (const auto& [map_addr, map_size] : page_stash) {
+        rasterizer->UnmapMemory(map_addr, map_size);
     }
+    page_stash.clear();
 
     BigPageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
     PageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
@@ -355,7 +360,7 @@ inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t si
     }
 }
 
-template <bool is_safe>
+template <bool is_safe, bool use_fastmem>
 void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
                                   [[maybe_unused]] VideoCommon::CacheType which) const {
     auto set_to_zero = [&]([[maybe_unused]] std::size_t page_index,
@@ -369,8 +374,12 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
         if constexpr (is_safe) {
             rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
         }
-        u8* physical = memory.GetPointer(cpu_addr_base);
-        std::memcpy(dest_buffer, physical, copy_amount);
+        if constexpr (use_fastmem) {
+            std::memcpy(dest_buffer, &fastmem_arena[cpu_addr_base], copy_amount);
+        } else {
+            u8* physical = memory.GetPointer(cpu_addr_base);
+            std::memcpy(dest_buffer, physical, copy_amount);
+        }
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
     };
     auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
@@ -379,11 +388,15 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
         if constexpr (is_safe) {
             rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
         }
-        if (!IsBigPageContinous(page_index)) [[unlikely]] {
-            memory.ReadBlockUnsafe(cpu_addr_base, dest_buffer, copy_amount);
+        if constexpr (use_fastmem) {
+            std::memcpy(dest_buffer, &fastmem_arena[cpu_addr_base], copy_amount);
         } else {
-            u8* physical = memory.GetPointer(cpu_addr_base);
-            std::memcpy(dest_buffer, physical, copy_amount);
+            if (!IsBigPageContinous(page_index)) [[unlikely]] {
+                memory.ReadBlockUnsafe(cpu_addr_base, dest_buffer, copy_amount);
+            } else {
+                u8* physical = memory.GetPointer(cpu_addr_base);
+                std::memcpy(dest_buffer, physical, copy_amount);
+            }
         }
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
     };
@@ -397,12 +410,20 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
 
 void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
                               VideoCommon::CacheType which) const {
-    ReadBlockImpl<true>(gpu_src_addr, dest_buffer, size, which);
+    if (fastmem_arena) [[likely]] {
+        ReadBlockImpl<true, true>(gpu_src_addr, dest_buffer, size, which);
+        return;
+    }
+    ReadBlockImpl<true, false>(gpu_src_addr, dest_buffer, size, which);
 }
 
 void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer,
                                     const std::size_t size) const {
-    ReadBlockImpl<false>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None);
+    if (fastmem_arena) [[likely]] {
+        ReadBlockImpl<false, true>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None);
+        return;
+    }
+    ReadBlockImpl<false, false>(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None);
 }
 
 template <bool is_safe>
@@ -454,6 +475,12 @@ void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buf
     WriteBlockImpl<false>(gpu_dest_addr, src_buffer, size, VideoCommon::CacheType::None);
 }
 
+void MemoryManager::WriteBlockCached(GPUVAddr gpu_dest_addr, const void* src_buffer,
+                                     std::size_t size) {
+    WriteBlockImpl<false>(gpu_dest_addr, src_buffer, size, VideoCommon::CacheType::None);
+    accumulator->Add(gpu_dest_addr, size);
+}
+
 void MemoryManager::FlushRegion(GPUVAddr gpu_addr, size_t size,
                                 VideoCommon::CacheType which) const {
     auto do_nothing = [&]([[maybe_unused]] std::size_t page_index,
@@ -663,7 +690,17 @@ bool MemoryManager::IsFullyMappedRange(GPUVAddr gpu_addr, std::size_t size) cons
 std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
     GPUVAddr gpu_addr, std::size_t size) const {
     std::vector<std::pair<GPUVAddr, std::size_t>> result{};
-    std::optional<std::pair<GPUVAddr, std::size_t>> last_segment{};
+    GetSubmappedRangeImpl<true>(gpu_addr, size, result);
+    return result;
+}
+
+template <bool is_gpu_address>
+void MemoryManager::GetSubmappedRangeImpl(
+    GPUVAddr gpu_addr, std::size_t size,
+    std::vector<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, VAddr>, std::size_t>>&
+        result) const {
+    std::optional<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, VAddr>, std::size_t>>
+        last_segment{};
     std::optional<VAddr> old_page_addr{};
     const auto split = [&last_segment, &result]([[maybe_unused]] std::size_t page_index,
                                                 [[maybe_unused]] std::size_t offset,
@@ -685,8 +722,12 @@ std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
         }
         old_page_addr = {cpu_addr_base + copy_amount};
         if (!last_segment) {
-            const GPUVAddr new_base_addr = (page_index << big_page_bits) + offset;
-            last_segment = {new_base_addr, copy_amount};
+            if constexpr (is_gpu_address) {
+                const GPUVAddr new_base_addr = (page_index << big_page_bits) + offset;
+                last_segment = {new_base_addr, copy_amount};
+            } else {
+                last_segment = {cpu_addr_base, copy_amount};
+            }
         } else {
             last_segment->second += copy_amount;
         }
@@ -703,8 +744,12 @@ std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
         }
         old_page_addr = {cpu_addr_base + copy_amount};
         if (!last_segment) {
-            const GPUVAddr new_base_addr = (page_index << page_bits) + offset;
-            last_segment = {new_base_addr, copy_amount};
+            if constexpr (is_gpu_address) {
+                const GPUVAddr new_base_addr = (page_index << page_bits) + offset;
+                last_segment = {new_base_addr, copy_amount};
+            } else {
+                last_segment = {cpu_addr_base, copy_amount};
+            }
         } else {
             last_segment->second += copy_amount;
         }
@@ -715,7 +760,18 @@ std::vector<std::pair<GPUVAddr, std::size_t>> MemoryManager::GetSubmappedRange(
     };
     MemoryOperation<true>(gpu_addr, size, extend_size_big, split, do_short_pages);
     split(0, 0, 0);
-    return result;
+}
+
+void MemoryManager::FlushCaching() {
+    if (!accumulator->AnyAccumulated()) {
+        return;
+    }
+    accumulator->Callback([this](GPUVAddr addr, size_t size) {
+        GetSubmappedRangeImpl<false>(addr, size, page_stash);
+    });
+    rasterizer->InnerInvalidation(page_stash);
+    page_stash.clear();
+    accumulator->Clear();
 }
 
 } // namespace Tegra

src/video_core/memory_manager.h

@@ -19,6 +19,10 @@ namespace VideoCore {
 class RasterizerInterface;
 }
 
+namespace VideoCommon {
+class InvalidationAccumulator;
+}
+
 namespace Core {
 class DeviceMemory;
 namespace Memory {
@@ -80,6 +84,7 @@ public:
      */
     void ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size) const;
     void WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size);
+    void WriteBlockCached(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size);
 
     /**
      * Checks if a gpu region can be simply read with a pointer.
@@ -129,12 +134,14 @@ public:
     size_t GetMemoryLayoutSize(GPUVAddr gpu_addr,
                                size_t max_size = std::numeric_limits<size_t>::max()) const;
 
+    void FlushCaching();
+
 private:
     template <bool is_big_pages, typename FuncMapped, typename FuncReserved, typename FuncUnmapped>
     inline void MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size, FuncMapped&& func_mapped,
                                 FuncReserved&& func_reserved, FuncUnmapped&& func_unmapped) const;
 
-    template <bool is_safe>
+    template <bool is_safe, bool use_fastmem>
     void ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
                        VideoCommon::CacheType which) const;
 
@@ -154,6 +161,12 @@ private:
     inline bool IsBigPageContinous(size_t big_page_index) const;
     inline void SetBigPageContinous(size_t big_page_index, bool value);
 
+    template <bool is_gpu_address>
+    void GetSubmappedRangeImpl(
+        GPUVAddr gpu_addr, std::size_t size,
+        std::vector<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, VAddr>, std::size_t>>&
+            result) const;
+
     Core::System& system;
     Core::Memory::Memory& memory;
     Core::DeviceMemory& device_memory;
@@ -201,10 +214,13 @@ private:
     Common::VirtualBuffer<u32> big_page_table_cpu;
 
     std::vector<u64> big_page_continous;
+    std::vector<std::pair<VAddr, std::size_t>> page_stash{};
+    u8* fastmem_arena{};
 
     constexpr static size_t continous_bits = 64;
 
     const size_t unique_identifier;
+    std::unique_ptr<VideoCommon::InvalidationAccumulator> accumulator;
 
     static std::atomic<size_t> unique_identifier_generator;
 };

src/video_core/rasterizer_interface.h

@@ -6,6 +6,7 @@
 #include <functional>
 #include <optional>
 #include <span>
+#include <utility>
 #include "common/common_types.h"
 #include "common/polyfill_thread.h"
 #include "video_core/cache_types.h"
@@ -95,6 +96,12 @@ public:
     virtual void InvalidateRegion(VAddr addr, u64 size,
                                   VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
+    virtual void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) {
+        for (const auto& [cpu_addr, size] : sequences) {
+            InvalidateRegion(cpu_addr, size);
+        }
+    }
+
     /// Notify rasterizer that any caches of the specified region are desync with guest
     virtual void OnCPUWrite(VAddr addr, u64 size) = 0;
 

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -160,6 +160,10 @@ public:
         return device.CanReportMemoryUsage();
     }
 
+    u32 GetStorageBufferAlignment() const {
+        return static_cast<u32>(device.GetShaderStorageBufferAlignment());
+    }
+
 private:
     static constexpr std::array PABO_LUT{
         GL_VERTEX_PROGRAM_PARAMETER_BUFFER_NV,          GL_TESS_CONTROL_PROGRAM_PARAMETER_BUFFER_NV,

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -139,6 +139,7 @@ void RasterizerOpenGL::LoadDiskResources(u64 title_id, std::stop_token stop_load
 void RasterizerOpenGL::Clear(u32 layer_count) {
     MICROPROFILE_SCOPE(OpenGL_Clears);
 
+    gpu_memory->FlushCaching();
     const auto& regs = maxwell3d->regs;
     bool use_color{};
     bool use_depth{};
@@ -207,6 +208,7 @@ void RasterizerOpenGL::PrepareDraw(bool is_indexed, Func&& draw_func) {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
+    gpu_memory->FlushCaching();
     query_cache.UpdateCounters();
 
     GraphicsPipeline* const pipeline{shader_cache.CurrentGraphicsPipeline()};
@@ -319,6 +321,7 @@ void RasterizerOpenGL::DrawIndirect() {
 }
 
 void RasterizerOpenGL::DispatchCompute() {
+    gpu_memory->FlushCaching();
     ComputePipeline* const pipeline{shader_cache.CurrentComputePipeline()};
     if (!pipeline) {
         return;
@@ -526,6 +529,7 @@ void RasterizerOpenGL::TickFrame() {
 }
 
 bool RasterizerOpenGL::AccelerateConditionalRendering() {
+    gpu_memory->FlushCaching();
     if (Settings::IsGPULevelHigh()) {
         // Reimplement Host conditional rendering.
         return false;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -236,6 +236,8 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .needs_demote_reorder = device.IsAmd(),
           .support_snorm_render_buffer = false,
           .support_viewport_index_layer = device.HasVertexViewportLayer(),
+          .min_ssbo_alignment = static_cast<u32>(device.GetShaderStorageBufferAlignment()),
+          .support_geometry_shader_passthrough = device.HasGeometryShaderPassthrough(),
       } {
     if (use_asynchronous_shaders) {
         workers = CreateWorkers();

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -442,7 +442,13 @@ void RendererOpenGL::DrawScreen(const Layout::FramebufferLayout& layout) {
 
     glBindTextureUnit(0, screen_info.display_texture);
 
-    const auto anti_aliasing = Settings::values.anti_aliasing.GetValue();
+    auto anti_aliasing = Settings::values.anti_aliasing.GetValue();
+    if (anti_aliasing > Settings::AntiAliasing::LastAA) {
+        LOG_ERROR(Render_OpenGL, "Invalid antialiasing option selected {}", anti_aliasing);
+        anti_aliasing = Settings::AntiAliasing::None;
+        Settings::values.anti_aliasing.SetValue(anti_aliasing);
+    }
+
     if (anti_aliasing != Settings::AntiAliasing::None) {
         glEnablei(GL_SCISSOR_TEST, 0);
         auto viewport_width = screen_info.texture.width;

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -60,24 +60,13 @@ std::string GetDriverVersion(const Device& device) {
     return GetReadableVersion(version);
 }
 
-std::string BuildCommaSeparatedExtensions(std::vector<std::string> available_extensions) {
-    std::sort(std::begin(available_extensions), std::end(available_extensions));
-
-    static constexpr std::size_t AverageExtensionSize = 64;
-    std::string separated_extensions;
-    separated_extensions.reserve(available_extensions.size() * AverageExtensionSize);
-
-    const auto end = std::end(available_extensions);
-    for (auto extension = std::begin(available_extensions); extension != end; ++extension) {
-        if (const bool is_last = extension + 1 == end; is_last) {
-            separated_extensions += *extension;
-        } else {
-            separated_extensions += fmt::format("{},", *extension);
-        }
-    }
-    return separated_extensions;
+std::string BuildCommaSeparatedExtensions(
+    const std::set<std::string, std::less<>>& available_extensions) {
+    return fmt::format("{}", fmt::join(available_extensions, ","));
 }
 
+} // Anonymous namespace
+
 Device CreateDevice(const vk::Instance& instance, const vk::InstanceDispatch& dld,
                     VkSurfaceKHR surface) {
     const std::vector<VkPhysicalDevice> devices = instance.EnumeratePhysicalDevices();
@@ -89,7 +78,6 @@ Device CreateDevice(const vk::Instance& instance, const vk::InstanceDispatch& dl
     const vk::PhysicalDevice physical_device(devices[device_index], dld);
     return Device(*instance, physical_device, surface, dld);
 }
-} // Anonymous namespace
 
 RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
                                Core::Frontend::EmuWindow& emu_window,
@@ -98,7 +86,7 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
     : RendererBase(emu_window, std::move(context_)), telemetry_session(telemetry_session_),
       cpu_memory(cpu_memory_), gpu(gpu_), library(OpenLibrary()),
       instance(CreateInstance(library, dld, VK_API_VERSION_1_1, render_window.GetWindowInfo().type,
-                              true, Settings::values.renderer_debug.GetValue())),
+                              Settings::values.renderer_debug.GetValue())),
       debug_callback(Settings::values.renderer_debug ? CreateDebugCallback(instance) : nullptr),
       surface(CreateSurface(instance, render_window)),
       device(CreateDevice(instance, dld, *surface)), memory_allocator(device, false),
@@ -109,6 +97,10 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
                   screen_info),
       rasterizer(render_window, gpu, cpu_memory, screen_info, device, memory_allocator,
                  state_tracker, scheduler) {
+    if (Settings::values.renderer_force_max_clock.GetValue() && device.ShouldBoostClocks()) {
+        turbo_mode.emplace(instance, dld);
+        scheduler.RegisterOnSubmit([this] { turbo_mode->QueueSubmitted(); });
+    }
     Report();
 } catch (const vk::Exception& exception) {
     LOG_ERROR(Render_Vulkan, "Vulkan initialization failed with error: {}", exception.what());
@@ -116,6 +108,7 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
 }
 
 RendererVulkan::~RendererVulkan() {
+    scheduler.RegisterOnSubmit([] {});
     void(device.GetLogical().WaitIdle());
 }
 

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -13,6 +13,7 @@
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_state_tracker.h"
 #include "video_core/renderer_vulkan/vk_swapchain.h"
+#include "video_core/renderer_vulkan/vk_turbo_mode.h"
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
@@ -31,6 +32,9 @@ class GPU;
 
 namespace Vulkan {
 
+Device CreateDevice(const vk::Instance& instance, const vk::InstanceDispatch& dld,
+                    VkSurfaceKHR surface);
+
 class RendererVulkan final : public VideoCore::RendererBase {
 public:
     explicit RendererVulkan(Core::TelemetrySession& telemtry_session,
@@ -74,6 +78,7 @@ private:
     Swapchain swapchain;
     BlitScreen blit_screen;
     RasterizerVulkan rasterizer;
+    std::optional<TurboMode> turbo_mode;
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -330,12 +330,19 @@ bool BufferCacheRuntime::CanReportMemoryUsage() const {
     return device.CanReportMemoryUsage();
 }
 
+u32 BufferCacheRuntime::GetStorageBufferAlignment() const {
+    return static_cast<u32>(device.GetStorageBufferAlignment());
+}
+
 void BufferCacheRuntime::Finish() {
     scheduler.Finish();
 }
 
 void BufferCacheRuntime::CopyBuffer(VkBuffer dst_buffer, VkBuffer src_buffer,
                                     std::span<const VideoCommon::BufferCopy> copies, bool barrier) {
+    if (dst_buffer == VK_NULL_HANDLE || src_buffer == VK_NULL_HANDLE) {
+        return;
+    }
     static constexpr VkMemoryBarrier READ_BARRIER{
         .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
         .pNext = nullptr,
@@ -394,6 +401,9 @@ void BufferCacheRuntime::PostCopyBarrier() {
 }
 
 void BufferCacheRuntime::ClearBuffer(VkBuffer dest_buffer, u32 offset, size_t size, u32 value) {
+    if (dest_buffer == VK_NULL_HANDLE) {
+        return;
+    }
     static constexpr VkMemoryBarrier READ_BARRIER{
         .sType = VK_STRUCTURE_TYPE_MEMORY_BARRIER,
         .pNext = nullptr,
@@ -473,6 +483,11 @@ void BufferCacheRuntime::BindVertexBuffer(u32 index, VkBuffer buffer, u32 offset
             cmdbuf.BindVertexBuffers2EXT(index, 1, &buffer, &vk_offset, &vk_size, &vk_stride);
         });
     } else {
+        if (!device.HasNullDescriptor() && buffer == VK_NULL_HANDLE) {
+            ReserveNullBuffer();
+            buffer = *null_buffer;
+            offset = 0;
+        }
         scheduler.Record([index, buffer, offset](vk::CommandBuffer cmdbuf) {
             cmdbuf.BindVertexBuffer(index, buffer, offset);
         });

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -73,6 +73,8 @@ public:
 
     bool CanReportMemoryUsage() const;
 
+    u32 GetStorageBufferAlignment() const;
+
     [[nodiscard]] StagingBufferRef UploadStagingBuffer(size_t size);
 
     [[nodiscard]] StagingBufferRef DownloadStagingBuffer(size_t size);

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -331,6 +331,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .need_declared_frag_colors = false,
 
         .has_broken_spirv_clamp = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS,
+        .has_broken_spirv_position_input = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY,
         .has_broken_unsigned_image_offsets = false,
         .has_broken_signed_operations = false,
         .has_broken_fp16_float_controls = driver_id == VK_DRIVER_ID_NVIDIA_PROPRIETARY,
@@ -343,6 +344,8 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
             driver_id == VK_DRIVER_ID_AMD_PROPRIETARY || driver_id == VK_DRIVER_ID_AMD_OPEN_SOURCE,
         .support_snorm_render_buffer = true,
         .support_viewport_index_layer = device.IsExtShaderViewportIndexLayerSupported(),
+        .min_ssbo_alignment = static_cast<u32>(device.GetStorageBufferAlignment()),
+        .support_geometry_shader_passthrough = device.IsNvGeometryShaderPassthroughSupported(),
     };
 
     if (device.GetMaxVertexInputAttributes() < Maxwell::NumVertexAttributes) {
@@ -790,7 +793,8 @@ vk::PipelineCache PipelineCache::LoadVulkanPipelineCache(const std::filesystem::
             return create_pipeline_cache(0, nullptr);
         }
 
-        const size_t cache_size = static_cast<size_t>(end) - magic_number.size();
+        static constexpr size_t header_size = magic_number.size() + sizeof(cache_version);
+        const size_t cache_size = static_cast<size_t>(end) - header_size;
         std::vector<char> cache_data(cache_size);
         file.read(cache_data.data(), cache_size);
 

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -186,6 +186,7 @@ void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
 
     SCOPE_EXIT({ gpu.TickWork(); });
     FlushWork();
+    gpu_memory->FlushCaching();
 
     query_cache.UpdateCounters();
 
@@ -269,6 +270,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
     MICROPROFILE_SCOPE(Vulkan_Clearing);
 
     FlushWork();
+    gpu_memory->FlushCaching();
 
     query_cache.UpdateCounters();
 
@@ -393,6 +395,7 @@ void RasterizerVulkan::Clear(u32 layer_count) {
 
 void RasterizerVulkan::DispatchCompute() {
     FlushWork();
+    gpu_memory->FlushCaching();
 
     ComputePipeline* const pipeline{pipeline_cache.CurrentComputePipeline()};
     if (!pipeline) {
@@ -481,6 +484,27 @@ void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size, VideoCommon::Cache
     }
 }
 
+void RasterizerVulkan::InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) {
+    {
+        std::scoped_lock lock{texture_cache.mutex};
+        for (const auto& [addr, size] : sequences) {
+            texture_cache.WriteMemory(addr, size);
+        }
+    }
+    {
+        std::scoped_lock lock{buffer_cache.mutex};
+        for (const auto& [addr, size] : sequences) {
+            buffer_cache.WriteMemory(addr, size);
+        }
+    }
+    {
+        for (const auto& [addr, size] : sequences) {
+            query_cache.InvalidateRegion(addr, size);
+            pipeline_cache.InvalidateRegion(addr, size);
+        }
+    }
+}
+
 void RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
     if (addr == 0 || size == 0) {
         return;
@@ -605,6 +629,7 @@ void RasterizerVulkan::TickFrame() {
 }
 
 bool RasterizerVulkan::AccelerateConditionalRendering() {
+    gpu_memory->FlushCaching();
     if (Settings::IsGPULevelHigh()) {
         // TODO(Blinkhawk): Reimplement Host conditional rendering.
         return false;

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -79,6 +79,7 @@ public:
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void InvalidateRegion(VAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+    void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) override;
     void OnCPUWrite(VAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
     void UnmapMemory(VAddr addr, u64 size) override;

src/video_core/renderer_vulkan/vk_scheduler.cpp

@@ -213,6 +213,11 @@ void Scheduler::SubmitExecution(VkSemaphore signal_semaphore, VkSemaphore wait_s
             .signalSemaphoreCount = num_signal_semaphores,
             .pSignalSemaphores = signal_semaphores.data(),
         };
+
+        if (on_submit) {
+            on_submit();
+        }
+
         switch (const VkResult result = device.GetGraphicsQueue().Submit(submit_info)) {
         case VK_SUCCESS:
             break;

src/video_core/renderer_vulkan/vk_scheduler.h

@@ -5,6 +5,7 @@
 
 #include <condition_variable>
 #include <cstddef>
+#include <functional>
 #include <memory>
 #include <thread>
 #include <utility>
@@ -66,6 +67,11 @@ public:
         query_cache = &query_cache_;
     }
 
+    // Registers a callback to perform on queue submission.
+    void RegisterOnSubmit(std::function<void()>&& func) {
+        on_submit = std::move(func);
+    }
+
     /// Send work to a separate thread.
     template <typename T>
     void Record(T&& command) {
@@ -216,6 +222,7 @@ private:
     vk::CommandBuffer current_cmdbuf;
 
     std::unique_ptr<CommandChunk> chunk;
+    std::function<void()> on_submit;
 
     State state;
 

src/video_core/renderer_vulkan/vk_turbo_mode.cpp

@@ -0,0 +1,222 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include "common/literals.h"
+#include "video_core/host_shaders/vulkan_turbo_mode_comp_spv.h"
+#include "video_core/renderer_vulkan/renderer_vulkan.h"
+#include "video_core/renderer_vulkan/vk_shader_util.h"
+#include "video_core/renderer_vulkan/vk_turbo_mode.h"
+#include "video_core/vulkan_common/vulkan_device.h"
+
+namespace Vulkan {
+
+using namespace Common::Literals;
+
+TurboMode::TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld)
+    : m_device{CreateDevice(instance, dld, VK_NULL_HANDLE)}, m_allocator{m_device, false} {
+    {
+        std::scoped_lock lk{m_submission_lock};
+        m_submission_time = std::chrono::steady_clock::now();
+    }
+    m_thread = std::jthread([&](auto stop_token) { Run(stop_token); });
+}
+
+TurboMode::~TurboMode() = default;
+
+void TurboMode::QueueSubmitted() {
+    std::scoped_lock lk{m_submission_lock};
+    m_submission_time = std::chrono::steady_clock::now();
+    m_submission_cv.notify_one();
+}
+
+void TurboMode::Run(std::stop_token stop_token) {
+    auto& dld = m_device.GetLogical();
+
+    // Allocate buffer. 2MiB should be sufficient.
+    auto buffer = dld.CreateBuffer(VkBufferCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .size = 2_MiB,
+        .usage = VK_BUFFER_USAGE_STORAGE_BUFFER_BIT | VK_BUFFER_USAGE_TRANSFER_DST_BIT,
+        .sharingMode = VK_SHARING_MODE_EXCLUSIVE,
+        .queueFamilyIndexCount = 0,
+        .pQueueFamilyIndices = nullptr,
+    });
+
+    // Commit some device local memory for the buffer.
+    auto commit = m_allocator.Commit(buffer, MemoryUsage::DeviceLocal);
+
+    // Create the descriptor pool to contain our descriptor.
+    constexpr VkDescriptorPoolSize pool_size{
+        .type = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+    };
+
+    auto descriptor_pool = dld.CreateDescriptorPool(VkDescriptorPoolCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_POOL_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = VK_DESCRIPTOR_POOL_CREATE_FREE_DESCRIPTOR_SET_BIT,
+        .maxSets = 1,
+        .poolSizeCount = 1,
+        .pPoolSizes = &pool_size,
+    });
+
+    // Create the descriptor set layout from the pool.
+    constexpr VkDescriptorSetLayoutBinding layout_binding{
+        .binding = 0,
+        .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+        .descriptorCount = 1,
+        .stageFlags = VK_SHADER_STAGE_COMPUTE_BIT,
+        .pImmutableSamplers = nullptr,
+    };
+
+    auto descriptor_set_layout = dld.CreateDescriptorSetLayout(VkDescriptorSetLayoutCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .bindingCount = 1,
+        .pBindings = &layout_binding,
+    });
+
+    // Actually create the descriptor set.
+    auto descriptor_set = descriptor_pool.Allocate(VkDescriptorSetAllocateInfo{
+        .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_ALLOCATE_INFO,
+        .pNext = nullptr,
+        .descriptorPool = *descriptor_pool,
+        .descriptorSetCount = 1,
+        .pSetLayouts = descriptor_set_layout.address(),
+    });
+
+    // Create the shader.
+    auto shader = BuildShader(m_device, VULKAN_TURBO_MODE_COMP_SPV);
+
+    // Create the pipeline layout.
+    auto pipeline_layout = dld.CreatePipelineLayout(VkPipelineLayoutCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .setLayoutCount = 1,
+        .pSetLayouts = descriptor_set_layout.address(),
+        .pushConstantRangeCount = 0,
+        .pPushConstantRanges = nullptr,
+    });
+
+    // Actually create the pipeline.
+    const VkPipelineShaderStageCreateInfo shader_stage{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stage = VK_SHADER_STAGE_COMPUTE_BIT,
+        .module = *shader,
+        .pName = "main",
+        .pSpecializationInfo = nullptr,
+    };
+
+    auto pipeline = dld.CreateComputePipeline(VkComputePipelineCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stage = shader_stage,
+        .layout = *pipeline_layout,
+        .basePipelineHandle = VK_NULL_HANDLE,
+        .basePipelineIndex = 0,
+    });
+
+    // Create a fence to wait on.
+    auto fence = dld.CreateFence(VkFenceCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_FENCE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+    });
+
+    // Create a command pool to allocate a command buffer from.
+    auto command_pool = dld.CreateCommandPool(VkCommandPoolCreateInfo{
+        .sType = VK_STRUCTURE_TYPE_COMMAND_POOL_CREATE_INFO,
+        .pNext = nullptr,
+        .flags =
+            VK_COMMAND_POOL_CREATE_TRANSIENT_BIT | VK_COMMAND_POOL_CREATE_RESET_COMMAND_BUFFER_BIT,
+        .queueFamilyIndex = m_device.GetGraphicsFamily(),
+    });
+
+    // Create a single command buffer.
+    auto cmdbufs = command_pool.Allocate(1, VK_COMMAND_BUFFER_LEVEL_PRIMARY);
+    auto cmdbuf = vk::CommandBuffer{cmdbufs[0], m_device.GetDispatchLoader()};
+
+    while (!stop_token.stop_requested()) {
+        // Reset the fence.
+        fence.Reset();
+
+        // Update descriptor set.
+        const VkDescriptorBufferInfo buffer_info{
+            .buffer = *buffer,
+            .offset = 0,
+            .range = VK_WHOLE_SIZE,
+        };
+
+        const VkWriteDescriptorSet buffer_write{
+            .sType = VK_STRUCTURE_TYPE_WRITE_DESCRIPTOR_SET,
+            .pNext = nullptr,
+            .dstSet = descriptor_set[0],
+            .dstBinding = 0,
+            .dstArrayElement = 0,
+            .descriptorCount = 1,
+            .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+            .pImageInfo = nullptr,
+            .pBufferInfo = &buffer_info,
+            .pTexelBufferView = nullptr,
+        };
+
+        dld.UpdateDescriptorSets(std::array{buffer_write}, {});
+
+        // Set up the command buffer.
+        cmdbuf.Begin(VkCommandBufferBeginInfo{
+            .sType = VK_STRUCTURE_TYPE_COMMAND_BUFFER_BEGIN_INFO,
+            .pNext = nullptr,
+            .flags = VK_COMMAND_BUFFER_USAGE_ONE_TIME_SUBMIT_BIT,
+            .pInheritanceInfo = nullptr,
+        });
+
+        // Clear the buffer.
+        cmdbuf.FillBuffer(*buffer, 0, VK_WHOLE_SIZE, 0);
+
+        // Bind descriptor set.
+        cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline_layout, 0,
+                                  descriptor_set, {});
+
+        // Bind the pipeline.
+        cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_COMPUTE, *pipeline);
+
+        // Dispatch.
+        cmdbuf.Dispatch(64, 64, 1);
+
+        // Finish.
+        cmdbuf.End();
+
+        const VkSubmitInfo submit_info{
+            .sType = VK_STRUCTURE_TYPE_SUBMIT_INFO,
+            .pNext = nullptr,
+            .waitSemaphoreCount = 0,
+            .pWaitSemaphores = nullptr,
+            .pWaitDstStageMask = nullptr,
+            .commandBufferCount = 1,
+            .pCommandBuffers = cmdbuf.address(),
+            .signalSemaphoreCount = 0,
+            .pSignalSemaphores = nullptr,
+        };
+
+        m_device.GetGraphicsQueue().Submit(std::array{submit_info}, *fence);
+
+        // Wait for completion.
+        fence.Wait();
+
+        // Wait for the next graphics queue submission if necessary.
+        std::unique_lock lk{m_submission_lock};
+        Common::CondvarWait(m_submission_cv, lk, stop_token, [this] {
+            return (std::chrono::steady_clock::now() - m_submission_time) <=
+                   std::chrono::milliseconds{100};
+        });
+    }
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_turbo_mode.h

@@ -0,0 +1,35 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <chrono>
+#include <mutex>
+
+#include "common/polyfill_thread.h"
+#include "video_core/vulkan_common/vulkan_device.h"
+#include "video_core/vulkan_common/vulkan_memory_allocator.h"
+#include "video_core/vulkan_common/vulkan_wrapper.h"
+
+namespace Vulkan {
+
+class TurboMode {
+public:
+    explicit TurboMode(const vk::Instance& instance, const vk::InstanceDispatch& dld);
+    ~TurboMode();
+
+    void QueueSubmitted();
+
+private:
+    void Run(std::stop_token stop_token);
+
+    Device m_device;
+    MemoryAllocator m_allocator;
+    std::mutex m_submission_lock;
+    std::condition_variable_any m_submission_cv;
+    std::chrono::time_point<std::chrono::steady_clock> m_submission_time{};
+
+    std::jthread m_thread;
+};
+
+} // namespace Vulkan

src/video_core/vulkan_common/vulkan_device.h

@@ -3,6 +3,7 @@
 
 #pragma once
 
+#include <set>
 #include <span>
 #include <string>
 #include <unordered_map>
@@ -11,6 +12,156 @@
 #include "common/common_types.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
+// Define all features which may be used by the implementation here.
+// Vulkan version in the macro describes the minimum version required for feature availability.
+// If the Vulkan version is lower than the required version, the named extension is required.
+#define FOR_EACH_VK_FEATURE_1_1(FEATURE)                                                           \
+    FEATURE(EXT, SubgroupSizeControl, SUBGROUP_SIZE_CONTROL, subgroup_size_control)                \
+    FEATURE(KHR, 16BitStorage, 16BIT_STORAGE, bit16_storage)                                       \
+    FEATURE(KHR, ShaderAtomicInt64, SHADER_ATOMIC_INT64, shader_atomic_int64)                      \
+    FEATURE(KHR, ShaderDrawParameters, SHADER_DRAW_PARAMETERS, shader_draw_parameters)             \
+    FEATURE(KHR, ShaderFloat16Int8, SHADER_FLOAT16_INT8, shader_float16_int8)                      \
+    FEATURE(KHR, UniformBufferStandardLayout, UNIFORM_BUFFER_STANDARD_LAYOUT,                      \
+            uniform_buffer_standard_layout)                                                        \
+    FEATURE(KHR, VariablePointer, VARIABLE_POINTERS, variable_pointer)
+
+#define FOR_EACH_VK_FEATURE_1_2(FEATURE)                                                           \
+    FEATURE(EXT, HostQueryReset, HOST_QUERY_RESET, host_query_reset)                               \
+    FEATURE(KHR, 8BitStorage, 8BIT_STORAGE, bit8_storage)                                          \
+    FEATURE(KHR, TimelineSemaphore, TIMELINE_SEMAPHORE, timeline_semaphore)
+
+#define FOR_EACH_VK_FEATURE_1_3(FEATURE)                                                           \
+    FEATURE(EXT, ShaderDemoteToHelperInvocation, SHADER_DEMOTE_TO_HELPER_INVOCATION,               \
+            shader_demote_to_helper_invocation)
+
+// Define all features which may be used by the implementation and require an extension here.
+#define FOR_EACH_VK_FEATURE_EXT(FEATURE)                                                           \
+    FEATURE(EXT, CustomBorderColor, CUSTOM_BORDER_COLOR, custom_border_color)                      \
+    FEATURE(EXT, DepthClipControl, DEPTH_CLIP_CONTROL, depth_clip_control)                         \
+    FEATURE(EXT, ExtendedDynamicState, EXTENDED_DYNAMIC_STATE, extended_dynamic_state)             \
+    FEATURE(EXT, ExtendedDynamicState2, EXTENDED_DYNAMIC_STATE_2, extended_dynamic_state2)         \
+    FEATURE(EXT, ExtendedDynamicState3, EXTENDED_DYNAMIC_STATE_3, extended_dynamic_state3)         \
+    FEATURE(EXT, IndexTypeUint8, INDEX_TYPE_UINT8, index_type_uint8)                               \
+    FEATURE(EXT, LineRasterization, LINE_RASTERIZATION, line_rasterization)                        \
+    FEATURE(EXT, PrimitiveTopologyListRestart, PRIMITIVE_TOPOLOGY_LIST_RESTART,                    \
+            primitive_topology_list_restart)                                                       \
+    FEATURE(EXT, ProvokingVertex, PROVOKING_VERTEX, provoking_vertex)                              \
+    FEATURE(EXT, Robustness2, ROBUSTNESS_2, robustness2)                                           \
+    FEATURE(EXT, TransformFeedback, TRANSFORM_FEEDBACK, transform_feedback)                        \
+    FEATURE(EXT, VertexInputDynamicState, VERTEX_INPUT_DYNAMIC_STATE, vertex_input_dynamic_state)  \
+    FEATURE(KHR, PipelineExecutableProperties, PIPELINE_EXECUTABLE_PROPERTIES,                     \
+            pipeline_executable_properties)                                                        \
+    FEATURE(KHR, WorkgroupMemoryExplicitLayout, WORKGROUP_MEMORY_EXPLICIT_LAYOUT,                  \
+            workgroup_memory_explicit_layout)
+
+// Define miscellaneous extensions which may be used by the implementation here.
+#define FOR_EACH_VK_EXTENSION(EXTENSION)                                                           \
+    EXTENSION(EXT, CONSERVATIVE_RASTERIZATION, conservative_rasterization)                         \
+    EXTENSION(EXT, DEPTH_RANGE_UNRESTRICTED, depth_range_unrestricted)                             \
+    EXTENSION(EXT, MEMORY_BUDGET, memory_budget)                                                   \
+    EXTENSION(EXT, ROBUSTNESS_2, robustness_2)                                                     \
+    EXTENSION(EXT, SAMPLER_FILTER_MINMAX, sampler_filter_minmax)                                   \
+    EXTENSION(EXT, SHADER_STENCIL_EXPORT, shader_stencil_export)                                   \
+    EXTENSION(EXT, SHADER_VIEWPORT_INDEX_LAYER, shader_viewport_index_layer)                       \
+    EXTENSION(EXT, TOOLING_INFO, tooling_info)                                                     \
+    EXTENSION(EXT, VERTEX_ATTRIBUTE_DIVISOR, vertex_attribute_divisor)                             \
+    EXTENSION(KHR, DRAW_INDIRECT_COUNT, draw_indirect_count)                                       \
+    EXTENSION(KHR, DRIVER_PROPERTIES, driver_properties)                                           \
+    EXTENSION(KHR, EXTERNAL_MEMORY_FD, external_memory_fd)                                         \
+    EXTENSION(KHR, PUSH_DESCRIPTOR, push_descriptor)                                               \
+    EXTENSION(KHR, SAMPLER_MIRROR_CLAMP_TO_EDGE, sampler_mirror_clamp_to_edge)                     \
+    EXTENSION(KHR, SHADER_FLOAT_CONTROLS, shader_float_controls)                                   \
+    EXTENSION(KHR, SPIRV_1_4, spirv_1_4)                                                           \
+    EXTENSION(KHR, SWAPCHAIN, swapchain)                                                           \
+    EXTENSION(KHR, SWAPCHAIN_MUTABLE_FORMAT, swapchain_mutable_format)                             \
+    EXTENSION(NV, DEVICE_DIAGNOSTICS_CONFIG, device_diagnostics_config)                            \
+    EXTENSION(NV, GEOMETRY_SHADER_PASSTHROUGH, geometry_shader_passthrough)                        \
+    EXTENSION(NV, VIEWPORT_ARRAY2, viewport_array2)                                                \
+    EXTENSION(NV, VIEWPORT_SWIZZLE, viewport_swizzle)
+
+#define FOR_EACH_VK_EXTENSION_WIN32(EXTENSION)                                                     \
+    EXTENSION(KHR, EXTERNAL_MEMORY_WIN32, external_memory_win32)
+
+// Define extensions which must be supported.
+#define FOR_EACH_VK_MANDATORY_EXTENSION(EXTENSION_NAME)                                            \
+    EXTENSION_NAME(VK_EXT_ROBUSTNESS_2_EXTENSION_NAME)                                             \
+    EXTENSION_NAME(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME)                                 \
+    EXTENSION_NAME(VK_KHR_DRIVER_PROPERTIES_EXTENSION_NAME)                                        \
+    EXTENSION_NAME(VK_KHR_SAMPLER_MIRROR_CLAMP_TO_EDGE_EXTENSION_NAME)                             \
+    EXTENSION_NAME(VK_KHR_SHADER_FLOAT_CONTROLS_EXTENSION_NAME)
+
+#define FOR_EACH_VK_MANDATORY_EXTENSION_GENERIC(EXTENSION_NAME)                                    \
+    EXTENSION_NAME(VK_KHR_EXTERNAL_MEMORY_FD_EXTENSION_NAME)
+
+#define FOR_EACH_VK_MANDATORY_EXTENSION_WIN32(EXTENSION_NAME)                                      \
+    EXTENSION_NAME(VK_KHR_EXTERNAL_MEMORY_WIN32_EXTENSION_NAME)
+
+// Define extensions where the absence of the extension may result in a degraded experience.
+#define FOR_EACH_VK_RECOMMENDED_EXTENSION(EXTENSION_NAME)                                          \
+    EXTENSION_NAME(VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME)                               \
+    EXTENSION_NAME(VK_EXT_DEPTH_RANGE_UNRESTRICTED_EXTENSION_NAME)                                 \
+    EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME)                                   \
+    EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME)                                 \
+    EXTENSION_NAME(VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME)                                 \
+    EXTENSION_NAME(VK_EXT_LINE_RASTERIZATION_EXTENSION_NAME)                                       \
+    EXTENSION_NAME(VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME)                               \
+    EXTENSION_NAME(VK_NV_GEOMETRY_SHADER_PASSTHROUGH_EXTENSION_NAME)                               \
+    EXTENSION_NAME(VK_NV_VIEWPORT_ARRAY2_EXTENSION_NAME)                                           \
+    EXTENSION_NAME(VK_NV_VIEWPORT_SWIZZLE_EXTENSION_NAME)
+
+// Define features which must be supported.
+#define FOR_EACH_VK_MANDATORY_FEATURE(FEATURE_NAME)                                                \
+    FEATURE_NAME(bit16_storage, storageBuffer16BitAccess)                                          \
+    FEATURE_NAME(bit16_storage, uniformAndStorageBuffer16BitAccess)                                \
+    FEATURE_NAME(bit8_storage, storageBuffer8BitAccess)                                            \
+    FEATURE_NAME(bit8_storage, uniformAndStorageBuffer8BitAccess)                                  \
+    FEATURE_NAME(features, depthBiasClamp)                                                         \
+    FEATURE_NAME(features, depthClamp)                                                             \
+    FEATURE_NAME(features, drawIndirectFirstInstance)                                              \
+    FEATURE_NAME(features, dualSrcBlend)                                                           \
+    FEATURE_NAME(features, fillModeNonSolid)                                                       \
+    FEATURE_NAME(features, fragmentStoresAndAtomics)                                               \
+    FEATURE_NAME(features, geometryShader)                                                         \
+    FEATURE_NAME(features, imageCubeArray)                                                         \
+    FEATURE_NAME(features, independentBlend)                                                       \
+    FEATURE_NAME(features, largePoints)                                                            \
+    FEATURE_NAME(features, logicOp)                                                                \
+    FEATURE_NAME(features, multiDrawIndirect)                                                      \
+    FEATURE_NAME(features, multiViewport)                                                          \
+    FEATURE_NAME(features, occlusionQueryPrecise)                                                  \
+    FEATURE_NAME(features, robustBufferAccess)                                                     \
+    FEATURE_NAME(features, samplerAnisotropy)                                                      \
+    FEATURE_NAME(features, sampleRateShading)                                                      \
+    FEATURE_NAME(features, shaderClipDistance)                                                     \
+    FEATURE_NAME(features, shaderCullDistance)                                                     \
+    FEATURE_NAME(features, shaderImageGatherExtended)                                              \
+    FEATURE_NAME(features, shaderStorageImageWriteWithoutFormat)                                   \
+    FEATURE_NAME(features, tessellationShader)                                                     \
+    FEATURE_NAME(features, vertexPipelineStoresAndAtomics)                                         \
+    FEATURE_NAME(features, wideLines)                                                              \
+    FEATURE_NAME(host_query_reset, hostQueryReset)                                                 \
+    FEATURE_NAME(robustness2, nullDescriptor)                                                      \
+    FEATURE_NAME(robustness2, robustBufferAccess2)                                                 \
+    FEATURE_NAME(robustness2, robustImageAccess2)                                                  \
+    FEATURE_NAME(shader_demote_to_helper_invocation, shaderDemoteToHelperInvocation)               \
+    FEATURE_NAME(shader_draw_parameters, shaderDrawParameters)                                     \
+    FEATURE_NAME(timeline_semaphore, timelineSemaphore)                                            \
+    FEATURE_NAME(variable_pointer, variablePointers)                                               \
+    FEATURE_NAME(variable_pointer, variablePointersStorageBuffer)
+
+// Define features where the absence of the feature may result in a degraded experience.
+#define FOR_EACH_VK_RECOMMENDED_FEATURE(FEATURE_NAME)                                              \
+    FEATURE_NAME(custom_border_color, customBorderColors)                                          \
+    FEATURE_NAME(extended_dynamic_state, extendedDynamicState)                                     \
+    FEATURE_NAME(index_type_uint8, indexTypeUint8)                                                 \
+    FEATURE_NAME(primitive_topology_list_restart, primitiveTopologyListRestart)                    \
+    FEATURE_NAME(provoking_vertex, provokingVertexLast)                                            \
+    FEATURE_NAME(shader_float16_int8, shaderFloat16)                                               \
+    FEATURE_NAME(shader_float16_int8, shaderInt8)                                                  \
+    FEATURE_NAME(transform_feedback, transformFeedback)                                            \
+    FEATURE_NAME(uniform_buffer_standard_layout, uniformBufferStandardLayout)                      \
+    FEATURE_NAME(vertex_input_dynamic_state, vertexInputDynamicState)
+
 namespace Vulkan {
 
 class NsightAftermathTracker;
@@ -88,67 +239,69 @@ public:
 
     /// Returns the current Vulkan API version provided in Vulkan-formatted version numbers.
     u32 ApiVersion() const {
-        return properties.apiVersion;
+        return properties.properties.apiVersion;
     }
 
     /// Returns the current driver version provided in Vulkan-formatted version numbers.
     u32 GetDriverVersion() const {
-        return properties.driverVersion;
+        return properties.properties.driverVersion;
     }
 
     /// Returns the device name.
     std::string_view GetModelName() const {
-        return properties.deviceName;
+        return properties.properties.deviceName;
     }
 
     /// Returns the driver ID.
     VkDriverIdKHR GetDriverID() const {
-        return driver_id;
+        return properties.driver.driverID;
     }
 
+    bool ShouldBoostClocks() const;
+
     /// Returns uniform buffer alignment requeriment.
     VkDeviceSize GetUniformBufferAlignment() const {
-        return properties.limits.minUniformBufferOffsetAlignment;
+        return properties.properties.limits.minUniformBufferOffsetAlignment;
     }
 
     /// Returns storage alignment requeriment.
     VkDeviceSize GetStorageBufferAlignment() const {
-        return properties.limits.minStorageBufferOffsetAlignment;
+        return properties.properties.limits.minStorageBufferOffsetAlignment;
     }
 
     /// Returns the maximum range for storage buffers.
     VkDeviceSize GetMaxStorageBufferRange() const {
-        return properties.limits.maxStorageBufferRange;
+        return properties.properties.limits.maxStorageBufferRange;
     }
 
     /// Returns the maximum size for push constants.
     VkDeviceSize GetMaxPushConstantsSize() const {
-        return properties.limits.maxPushConstantsSize;
+        return properties.properties.limits.maxPushConstantsSize;
     }
 
     /// Returns the maximum size for shared memory.
     u32 GetMaxComputeSharedMemorySize() const {
-        return properties.limits.maxComputeSharedMemorySize;
+        return properties.properties.limits.maxComputeSharedMemorySize;
     }
 
     /// Returns float control properties of the device.
     const VkPhysicalDeviceFloatControlsPropertiesKHR& FloatControlProperties() const {
-        return float_controls;
+        return properties.float_controls;
     }
 
     /// Returns true if ASTC is natively supported.
     bool IsOptimalAstcSupported() const {
-        return is_optimal_astc_supported;
+        return features.features.textureCompressionASTC_LDR;
     }
 
     /// Returns true if the device supports float16 natively.
     bool IsFloat16Supported() const {
-        return is_float16_supported;
+        return features.shader_float16_int8.shaderFloat16;
     }
 
     /// Returns true if the device supports int8 natively.
     bool IsInt8Supported() const {
-        return is_int8_supported;
+        return features.shader_float16_int8.shaderInt8;
     }
 
     /// Returns true if the device warp size can potentially be bigger than guest's warp size.
@@ -158,32 +311,32 @@ public:
 
     /// Returns true if the device can be forced to use the guest warp size.
     bool IsGuestWarpSizeSupported(VkShaderStageFlagBits stage) const {
-        return guest_warp_stages & stage;
+        return properties.subgroup_size_control.requiredSubgroupSizeStages & stage;
     }
 
     /// Returns the maximum number of push descriptors.
     u32 MaxPushDescriptors() const {
-        return max_push_descriptors;
+        return properties.push_descriptor.maxPushDescriptors;
     }
 
     /// Returns true if formatless image load is supported.
     bool IsFormatlessImageLoadSupported() const {
-        return is_formatless_image_load_supported;
+        return features.features.shaderStorageImageReadWithoutFormat;
     }
 
     /// Returns true if shader int64 is supported.
     bool IsShaderInt64Supported() const {
-        return is_shader_int64_supported;
+        return features.features.shaderInt64;
     }
 
     /// Returns true if shader int16 is supported.
     bool IsShaderInt16Supported() const {
-        return is_shader_int16_supported;
+        return features.features.shaderInt16;
     }
 
     // Returns true if depth bounds is supported.
     bool IsDepthBoundsSupported() const {
-        return is_depth_bounds_supported;
+        return features.features.depthBounds;
     }
 
     /// Returns true when blitting from and to depth stencil images is supported.
@@ -193,151 +346,151 @@ public:
 
     /// Returns true if the device supports VK_NV_viewport_swizzle.
     bool IsNvViewportSwizzleSupported() const {
-        return nv_viewport_swizzle;
+        return extensions.viewport_swizzle;
     }
 
     /// Returns true if the device supports VK_NV_viewport_array2.
     bool IsNvViewportArray2Supported() const {
-        return nv_viewport_array2;
+        return extensions.viewport_array2;
     }
 
     /// Returns true if the device supports VK_NV_geometry_shader_passthrough.
     bool IsNvGeometryShaderPassthroughSupported() const {
-        return nv_geometry_shader_passthrough;
+        return extensions.geometry_shader_passthrough;
     }
 
     /// Returns true if the device supports VK_KHR_uniform_buffer_standard_layout.
     bool IsKhrUniformBufferStandardLayoutSupported() const {
-        return khr_uniform_buffer_standard_layout;
+        return extensions.uniform_buffer_standard_layout;
     }
 
     /// Returns true if the device supports VK_KHR_push_descriptor.
     bool IsKhrPushDescriptorSupported() const {
-        return khr_push_descriptor;
+        return extensions.push_descriptor;
     }
 
     /// Returns true if VK_KHR_pipeline_executable_properties is enabled.
     bool IsKhrPipelineExecutablePropertiesEnabled() const {
-        return khr_pipeline_executable_properties;
+        return extensions.pipeline_executable_properties;
     }
 
     /// Returns true if VK_KHR_swapchain_mutable_format is enabled.
     bool IsKhrSwapchainMutableFormatEnabled() const {
-        return khr_swapchain_mutable_format;
+        return extensions.swapchain_mutable_format;
     }
 
     /// Returns true if the device supports VK_KHR_workgroup_memory_explicit_layout.
     bool IsKhrWorkgroupMemoryExplicitLayoutSupported() const {
-        return khr_workgroup_memory_explicit_layout;
+        return extensions.workgroup_memory_explicit_layout;
     }
 
     /// Returns true if the device supports VK_EXT_primitive_topology_list_restart.
     bool IsTopologyListPrimitiveRestartSupported() const {
-        return is_topology_list_restart_supported;
+        return features.primitive_topology_list_restart.primitiveTopologyListRestart;
     }
 
     /// Returns true if the device supports VK_EXT_primitive_topology_list_restart.
     bool IsPatchListPrimitiveRestartSupported() const {
-        return is_patch_list_restart_supported;
+        return features.primitive_topology_list_restart.primitiveTopologyPatchListRestart;
     }
 
     /// Returns true if the device supports VK_EXT_index_type_uint8.
     bool IsExtIndexTypeUint8Supported() const {
-        return ext_index_type_uint8;
+        return extensions.index_type_uint8;
     }
 
     /// Returns true if the device supports VK_EXT_sampler_filter_minmax.
     bool IsExtSamplerFilterMinmaxSupported() const {
-        return ext_sampler_filter_minmax;
+        return extensions.sampler_filter_minmax;
     }
 
     /// Returns true if the device supports VK_EXT_depth_range_unrestricted.
     bool IsExtDepthRangeUnrestrictedSupported() const {
-        return ext_depth_range_unrestricted;
+        return extensions.depth_range_unrestricted;
     }
 
     /// Returns true if the device supports VK_EXT_depth_clip_control.
     bool IsExtDepthClipControlSupported() const {
-        return ext_depth_clip_control;
+        return extensions.depth_clip_control;
     }
 
     /// Returns true if the device supports VK_EXT_shader_viewport_index_layer.
     bool IsExtShaderViewportIndexLayerSupported() const {
-        return ext_shader_viewport_index_layer;
+        return extensions.shader_viewport_index_layer;
     }
 
     /// Returns true if the device supports VK_EXT_subgroup_size_control.
     bool IsExtSubgroupSizeControlSupported() const {
-        return ext_subgroup_size_control;
+        return extensions.subgroup_size_control;
     }
 
     /// Returns true if the device supports VK_EXT_transform_feedback.
     bool IsExtTransformFeedbackSupported() const {
-        return ext_transform_feedback;
+        return extensions.transform_feedback;
     }
 
     /// Returns true if the device supports VK_EXT_custom_border_color.
     bool IsExtCustomBorderColorSupported() const {
-        return ext_custom_border_color;
+        return extensions.custom_border_color;
     }
 
     /// Returns true if the device supports VK_EXT_extended_dynamic_state.
     bool IsExtExtendedDynamicStateSupported() const {
-        return ext_extended_dynamic_state;
+        return extensions.extended_dynamic_state;
     }
 
     /// Returns true if the device supports VK_EXT_extended_dynamic_state2.
     bool IsExtExtendedDynamicState2Supported() const {
-        return ext_extended_dynamic_state_2;
+        return extensions.extended_dynamic_state2;
     }
 
     bool IsExtExtendedDynamicState2ExtrasSupported() const {
-        return ext_extended_dynamic_state_2_extra;
+        return features.extended_dynamic_state2.extendedDynamicState2LogicOp;
     }
 
     /// Returns true if the device supports VK_EXT_extended_dynamic_state3.
     bool IsExtExtendedDynamicState3Supported() const {
-        return ext_extended_dynamic_state_3;
+        return extensions.extended_dynamic_state3;
     }
 
     /// Returns true if the device supports VK_EXT_extended_dynamic_state3.
     bool IsExtExtendedDynamicState3BlendingSupported() const {
-        return ext_extended_dynamic_state_3_blend;
+        return dynamic_state3_blending;
     }
 
     /// Returns true if the device supports VK_EXT_extended_dynamic_state3.
     bool IsExtExtendedDynamicState3EnablesSupported() const {
-        return ext_extended_dynamic_state_3_enables;
+        return dynamic_state3_enables;
     }
 
     /// Returns true if the device supports VK_EXT_line_rasterization.
     bool IsExtLineRasterizationSupported() const {
-        return ext_line_rasterization;
+        return extensions.line_rasterization;
     }
 
     /// Returns true if the device supports VK_EXT_vertex_input_dynamic_state.
     bool IsExtVertexInputDynamicStateSupported() const {
-        return ext_vertex_input_dynamic_state;
+        return extensions.vertex_input_dynamic_state;
     }
 
     /// Returns true if the device supports VK_EXT_shader_stencil_export.
     bool IsExtShaderStencilExportSupported() const {
-        return ext_shader_stencil_export;
+        return extensions.shader_stencil_export;
     }
 
     /// Returns true if the device supports VK_EXT_conservative_rasterization.
     bool IsExtConservativeRasterizationSupported() const {
-        return ext_conservative_rasterization;
+        return extensions.conservative_rasterization;
     }
 
     /// Returns true if the device supports VK_EXT_provoking_vertex.
     bool IsExtProvokingVertexSupported() const {
-        return ext_provoking_vertex;
+        return extensions.provoking_vertex;
     }
 
     /// Returns true if the device supports VK_KHR_shader_atomic_int64.
     bool IsExtShaderAtomicInt64Supported() const {
-        return ext_shader_atomic_int64;
+        return extensions.shader_atomic_int64;
     }
 
     /// Returns the minimum supported version of SPIR-V.
@@ -345,7 +498,7 @@ public:
         if (instance_version >= VK_API_VERSION_1_3) {
             return 0x00010600U;
         }
-        if (khr_spirv_1_4) {
+        if (extensions.spirv_1_4) {
             return 0x00010400U;
         }
         return 0x00010000U;
@@ -363,11 +516,11 @@ public:
 
     /// Returns the vendor name reported from Vulkan.
     std::string_view GetVendorName() const {
-        return vendor_name;
+        return properties.driver.driverName;
     }
 
     /// Returns the list of available extensions.
-    const std::vector<std::string>& GetAvailableExtensions() const {
+    const std::set<std::string, std::less<>>& GetAvailableExtensions() const {
         return supported_extensions;
     }
 
@@ -376,7 +529,7 @@ public:
     }
 
     bool CanReportMemoryUsage() const {
-        return ext_memory_budget;
+        return extensions.memory_budget;
     }
 
     u64 GetDeviceMemoryUsage() const;
@@ -397,33 +550,30 @@ public:
         return must_emulate_bgr565;
     }
 
+    bool HasNullDescriptor() const {
+        return features.robustness2.nullDescriptor;
+    }
+
     u32 GetMaxVertexInputAttributes() const {
-        return max_vertex_input_attributes;
+        return properties.properties.limits.maxVertexInputAttributes;
     }
 
     u32 GetMaxVertexInputBindings() const {
-        return max_vertex_input_bindings;
+        return properties.properties.limits.maxVertexInputBindings;
     }
 
 private:
-    /// Checks if the physical device is suitable.
-    void CheckSuitability(bool requires_swapchain) const;
+    /// Checks if the physical device is suitable and configures the object state
+    /// with all necessary info about its properties.
+    bool GetSuitability(bool requires_swapchain);
 
-    /// Loads extensions into a vector and stores available ones in this object.
-    std::vector<const char*> LoadExtensions(bool requires_surface);
+    // Remove extensions which have incomplete feature support.
+    void RemoveUnsuitableExtensions();
+    void RemoveExtensionIfUnsuitable(bool is_suitable, const std::string& extension_name);
 
     /// Sets up queue families.
     void SetupFamilies(VkSurfaceKHR surface);
 
-    /// Sets up device features.
-    void SetupFeatures();
-
-    /// Sets up device properties.
-    void SetupProperties();
-
-    /// Collects telemetry information from the device.
-    void CollectTelemetryParameters();
-
     /// Collects information about attached tools.
     void CollectToolingInfo();
 
@@ -434,90 +584,93 @@ private:
     std::vector<VkDeviceQueueCreateInfo> GetDeviceQueueCreateInfos() const;
 
     /// Returns true if ASTC textures are natively supported.
-    bool IsOptimalAstcSupported(const VkPhysicalDeviceFeatures& features) const;
+    bool ComputeIsOptimalAstcSupported() const;
 
     /// Returns true if the device natively supports blitting depth stencil images.
     bool TestDepthStencilBlits() const;
 
-    VkInstance instance;                                         ///< Vulkan instance.
-    vk::DeviceDispatch dld;                                      ///< Device function pointers.
-    vk::PhysicalDevice physical;                                 ///< Physical device.
-    VkPhysicalDeviceProperties properties;                       ///< Device properties.
-    VkPhysicalDeviceFloatControlsPropertiesKHR float_controls{}; ///< Float control properties.
-    vk::Device logical;                                          ///< Logical device.
-    vk::Queue graphics_queue;                                    ///< Main graphics queue.
-    vk::Queue present_queue;                                     ///< Main present queue.
-    u32 instance_version{};                                      ///< Vulkan onstance version.
-    u32 graphics_family{};                      ///< Main graphics queue family index.
-    u32 present_family{};                       ///< Main present queue family index.
-    VkDriverIdKHR driver_id{};                  ///< Driver ID.
-    VkShaderStageFlags guest_warp_stages{};     ///< Stages where the guest warp size can be forced.
-    u64 device_access_memory{};                 ///< Total size of device local memory in bytes.
-    u32 max_push_descriptors{};                 ///< Maximum number of push descriptors
-    u32 sets_per_pool{};                        ///< Sets per Description Pool
-    bool is_optimal_astc_supported{};           ///< Support for native ASTC.
-    bool is_float16_supported{};                ///< Support for float16 arithmetic.
-    bool is_int8_supported{};                   ///< Support for int8 arithmetic.
-    bool is_warp_potentially_bigger{};          ///< Host warp size can be bigger than guest.
-    bool is_formatless_image_load_supported{};  ///< Support for shader image read without format.
-    bool is_depth_bounds_supported{};           ///< Support for depth bounds.
-    bool is_shader_float64_supported{};         ///< Support for float64.
-    bool is_shader_int64_supported{};           ///< Support for int64.
-    bool is_shader_int16_supported{};           ///< Support for int16.
-    bool is_shader_storage_image_multisample{}; ///< Support for image operations on MSAA images.
-    bool is_blit_depth_stencil_supported{};     ///< Support for blitting from and to depth stencil.
-    bool is_topology_list_restart_supported{};  ///< Support for primitive restart with list
-                                                ///< topologies.
-    bool is_patch_list_restart_supported{};     ///< Support for primitive restart with list patch.
-    bool is_integrated{};                       ///< Is GPU an iGPU.
-    bool is_virtual{};                          ///< Is GPU a virtual GPU.
-    bool is_non_gpu{};                          ///< Is SoftwareRasterizer, FPGA, non-GPU device.
-    bool nv_viewport_swizzle{};                 ///< Support for VK_NV_viewport_swizzle.
-    bool nv_viewport_array2{};                  ///< Support for VK_NV_viewport_array2.
-    bool nv_geometry_shader_passthrough{};      ///< Support for VK_NV_geometry_shader_passthrough.
-    bool khr_draw_indirect_count{};             ///< Support for VK_KHR_draw_indirect_count.
-    bool khr_uniform_buffer_standard_layout{};  ///< Support for scalar uniform buffer layouts.
-    bool khr_spirv_1_4{};                       ///< Support for VK_KHR_spirv_1_4.
-    bool khr_workgroup_memory_explicit_layout{}; ///< Support for explicit workgroup layouts.
-    bool khr_push_descriptor{};                  ///< Support for VK_KHR_push_descritor.
-    bool khr_pipeline_executable_properties{};   ///< Support for executable properties.
-    bool khr_swapchain_mutable_format{};         ///< Support for VK_KHR_swapchain_mutable_format.
-    bool ext_index_type_uint8{};                 ///< Support for VK_EXT_index_type_uint8.
-    bool ext_sampler_filter_minmax{};            ///< Support for VK_EXT_sampler_filter_minmax.
-    bool ext_depth_clip_control{};               ///< Support for VK_EXT_depth_clip_control
-    bool ext_depth_range_unrestricted{};         ///< Support for VK_EXT_depth_range_unrestricted.
-    bool ext_shader_viewport_index_layer{};    ///< Support for VK_EXT_shader_viewport_index_layer.
-    bool ext_tooling_info{};                   ///< Support for VK_EXT_tooling_info.
-    bool ext_subgroup_size_control{};          ///< Support for VK_EXT_subgroup_size_control.
-    bool ext_transform_feedback{};             ///< Support for VK_EXT_transform_feedback.
-    bool ext_custom_border_color{};            ///< Support for VK_EXT_custom_border_color.
-    bool ext_extended_dynamic_state{};         ///< Support for VK_EXT_extended_dynamic_state.
-    bool ext_extended_dynamic_state_2{};       ///< Support for VK_EXT_extended_dynamic_state2.
-    bool ext_extended_dynamic_state_2_extra{}; ///< Support for VK_EXT_extended_dynamic_state2.
-    bool ext_extended_dynamic_state_3{};       ///< Support for VK_EXT_extended_dynamic_state3.
-    bool ext_extended_dynamic_state_3_blend{}; ///< Support for VK_EXT_extended_dynamic_state3.
-    bool ext_extended_dynamic_state_3_enables{}; ///< Support for VK_EXT_extended_dynamic_state3.
-    bool ext_line_rasterization{};               ///< Support for VK_EXT_line_rasterization.
-    bool ext_vertex_input_dynamic_state{};       ///< Support for VK_EXT_vertex_input_dynamic_state.
-    bool ext_shader_stencil_export{};            ///< Support for VK_EXT_shader_stencil_export.
-    bool ext_shader_atomic_int64{};              ///< Support for VK_KHR_shader_atomic_int64.
-    bool ext_conservative_rasterization{};       ///< Support for VK_EXT_conservative_rasterization.
-    bool ext_provoking_vertex{};                 ///< Support for VK_EXT_provoking_vertex.
-    bool ext_memory_budget{};                    ///< Support for VK_EXT_memory_budget.
-    bool nv_device_diagnostics_config{};         ///< Support for VK_NV_device_diagnostics_config.
-    bool has_broken_cube_compatibility{};        ///< Has broken cube compatiblity bit
-    bool has_renderdoc{};                        ///< Has RenderDoc attached
-    bool has_nsight_graphics{};                  ///< Has Nsight Graphics attached
-    bool supports_d24_depth{};                   ///< Supports D24 depth buffers.
-    bool cant_blit_msaa{};                       ///< Does not support MSAA<->MSAA blitting.
-    bool must_emulate_bgr565{};                  ///< Emulates BGR565 by swizzling RGB565 format.
-    u32 max_vertex_input_attributes{};           ///< Max vertex input attributes in pipeline
-    u32 max_vertex_input_bindings{};             ///< Max vertex input buffers in pipeline
+private:
+    VkInstance instance;         ///< Vulkan instance.
+    vk::DeviceDispatch dld;      ///< Device function pointers.
+    vk::PhysicalDevice physical; ///< Physical device.
+    vk::Device logical;          ///< Logical device.
+    vk::Queue graphics_queue;    ///< Main graphics queue.
+    vk::Queue present_queue;     ///< Main present queue.
+    u32 instance_version{};      ///< Vulkan instance version.
+    u32 graphics_family{};       ///< Main graphics queue family index.
+    u32 present_family{};        ///< Main present queue family index.
+
+    struct Extensions {
+#define EXTENSION(prefix, macro_name, var_name) bool var_name{};
+#define FEATURE(prefix, struct_name, macro_name, var_name) bool var_name{};
+
+        FOR_EACH_VK_FEATURE_1_1(FEATURE);
+        FOR_EACH_VK_FEATURE_1_2(FEATURE);
+        FOR_EACH_VK_FEATURE_1_3(FEATURE);
+        FOR_EACH_VK_FEATURE_EXT(FEATURE);
+        FOR_EACH_VK_EXTENSION(EXTENSION);
+        FOR_EACH_VK_EXTENSION_WIN32(EXTENSION);
+
+#undef EXTENSION
+#undef FEATURE
+    };
+
+    struct Features {
+#define FEATURE_CORE(prefix, struct_name, macro_name, var_name)                                    \
+    VkPhysicalDevice##struct_name##Features var_name{};
+#define FEATURE_EXT(prefix, struct_name, macro_name, var_name)                                     \
+    VkPhysicalDevice##struct_name##Features##prefix var_name{};
+
+        FOR_EACH_VK_FEATURE_1_1(FEATURE_CORE);
+        FOR_EACH_VK_FEATURE_1_2(FEATURE_CORE);
+        FOR_EACH_VK_FEATURE_1_3(FEATURE_CORE);
+        FOR_EACH_VK_FEATURE_EXT(FEATURE_EXT);
+
+#undef FEATURE_CORE
+#undef FEATURE_EXT
+
+        VkPhysicalDeviceFeatures features{};
+    };
+
+    struct Properties {
+        VkPhysicalDeviceDriverProperties driver{};
+        VkPhysicalDeviceFloatControlsProperties float_controls{};
+        VkPhysicalDevicePushDescriptorPropertiesKHR push_descriptor{};
+        VkPhysicalDeviceSubgroupSizeControlProperties subgroup_size_control{};
+        VkPhysicalDeviceTransformFeedbackPropertiesEXT transform_feedback{};
+
+        VkPhysicalDeviceProperties properties{};
+    };
+
+    Extensions extensions{};
+    Features features{};
+    Properties properties{};
+
+    VkPhysicalDeviceFeatures2 features2{};
+    VkPhysicalDeviceProperties2 properties2{};
+
+    // Misc features
+    bool is_optimal_astc_supported{};       ///< Support for all guest ASTC formats.
+    bool is_blit_depth_stencil_supported{}; ///< Support for blitting from and to depth stencil.
+    bool is_warp_potentially_bigger{};      ///< Host warp size can be bigger than guest.
+    bool is_integrated{};                   ///< Is GPU an iGPU.
+    bool is_virtual{};                      ///< Is GPU a virtual GPU.
+    bool is_non_gpu{};                      ///< Is SoftwareRasterizer, FPGA, non-GPU device.
+    bool has_broken_cube_compatibility{};   ///< Has broken cube compatiblity bit
+    bool has_renderdoc{};                   ///< Has RenderDoc attached
+    bool has_nsight_graphics{};             ///< Has Nsight Graphics attached
+    bool supports_d24_depth{};              ///< Supports D24 depth buffers.
+    bool cant_blit_msaa{};                  ///< Does not support MSAA<->MSAA blitting.
+    bool must_emulate_bgr565{};             ///< Emulates BGR565 by swizzling RGB565 format.
+    bool dynamic_state3_blending{};         ///< Has all blending features of dynamic_state3.
+    bool dynamic_state3_enables{};          ///< Has all enables features of dynamic_state3.
+    u64 device_access_memory{};             ///< Total size of device local memory in bytes.
+    u32 sets_per_pool{};                    ///< Sets per Description Pool
 
     // Telemetry parameters
-    std::string vendor_name;                       ///< Device's driver name.
-    std::vector<std::string> supported_extensions; ///< Reported Vulkan extensions.
-    std::vector<size_t> valid_heap_memory;         ///< Heaps used.
+    std::set<std::string, std::less<>> supported_extensions; ///< Reported Vulkan extensions.
+    std::set<std::string, std::less<>> loaded_extensions;    ///< Loaded Vulkan extensions.
+    std::vector<size_t> valid_heap_memory;                   ///< Heaps used.
 
     /// Format properties dictionary.
     std::unordered_map<VkFormat, VkFormatProperties> format_properties;

src/video_core/vulkan_common/vulkan_instance.cpp

@@ -32,7 +32,7 @@
 namespace Vulkan {
 namespace {
 [[nodiscard]] std::vector<const char*> RequiredExtensions(
-    Core::Frontend::WindowSystemType window_type, bool enable_debug_utils) {
+    Core::Frontend::WindowSystemType window_type, bool enable_validation) {
     std::vector<const char*> extensions;
     extensions.reserve(6);
     switch (window_type) {
@@ -65,7 +65,7 @@ namespace {
     if (window_type != Core::Frontend::WindowSystemType::Headless) {
         extensions.push_back(VK_KHR_SURFACE_EXTENSION_NAME);
     }
-    if (enable_debug_utils) {
+    if (enable_validation) {
         extensions.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
     }
     extensions.push_back(VK_KHR_GET_PHYSICAL_DEVICE_PROPERTIES_2_EXTENSION_NAME);
@@ -95,9 +95,9 @@ namespace {
     return true;
 }
 
-[[nodiscard]] std::vector<const char*> Layers(bool enable_layers) {
+[[nodiscard]] std::vector<const char*> Layers(bool enable_validation) {
     std::vector<const char*> layers;
-    if (enable_layers) {
+    if (enable_validation) {
         layers.push_back("VK_LAYER_KHRONOS_validation");
     }
     return layers;
@@ -125,7 +125,7 @@ void RemoveUnavailableLayers(const vk::InstanceDispatch& dld, std::vector<const
 
 vk::Instance CreateInstance(const Common::DynamicLibrary& library, vk::InstanceDispatch& dld,
                             u32 required_version, Core::Frontend::WindowSystemType window_type,
-                            bool enable_debug_utils, bool enable_layers) {
+                            bool enable_validation) {
     if (!library.IsOpen()) {
         LOG_ERROR(Render_Vulkan, "Vulkan library not available");
         throw vk::Exception(VK_ERROR_INITIALIZATION_FAILED);
@@ -138,11 +138,11 @@ vk::Instance CreateInstance(const Common::DynamicLibrary& library, vk::InstanceD
         LOG_ERROR(Render_Vulkan, "Failed to load Vulkan function pointers");
         throw vk::Exception(VK_ERROR_INITIALIZATION_FAILED);
     }
-    const std::vector<const char*> extensions = RequiredExtensions(window_type, enable_debug_utils);
+    const std::vector<const char*> extensions = RequiredExtensions(window_type, enable_validation);
     if (!AreExtensionsSupported(dld, extensions)) {
         throw vk::Exception(VK_ERROR_EXTENSION_NOT_PRESENT);
     }
-    std::vector<const char*> layers = Layers(enable_layers);
+    std::vector<const char*> layers = Layers(enable_validation);
     RemoveUnavailableLayers(dld, layers);
 
     const u32 available_version = vk::AvailableVersion(dld);

src/video_core/vulkan_common/vulkan_instance.h

@@ -17,8 +17,7 @@ namespace Vulkan {
  * @param dld                Dispatch table to load function pointers into
  * @param required_version   Required Vulkan version (for example, VK_API_VERSION_1_1)
  * @param window_type        Window system type's enabled extension
- * @param enable_debug_utils Whether to enable VK_EXT_debug_utils_extension_name or not
- * @param enable_layers      Whether to enable Vulkan validation layers or not
+ * @param enable_validation  Whether to enable Vulkan validation layers or not
  *
  * @return A new Vulkan instance
  * @throw vk::Exception on failure
@@ -26,6 +25,6 @@ namespace Vulkan {
 [[nodiscard]] vk::Instance CreateInstance(
     const Common::DynamicLibrary& library, vk::InstanceDispatch& dld, u32 required_version,
     Core::Frontend::WindowSystemType window_type = Core::Frontend::WindowSystemType::Headless,
-    bool enable_debug_utils = false, bool enable_layers = false);
+    bool enable_validation = false);
 
 } // namespace Vulkan

src/video_core/vulkan_common/vulkan_wrapper.cpp

@@ -96,8 +96,8 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkCmdDrawIndexed);
     X(vkCmdDrawIndirect);
     X(vkCmdDrawIndexedIndirect);
-    X(vkCmdDrawIndirectCountKHR);
-    X(vkCmdDrawIndexedIndirectCountKHR);
+    X(vkCmdDrawIndirectCount);
+    X(vkCmdDrawIndexedIndirectCount);
     X(vkCmdEndQuery);
     X(vkCmdEndRenderPass);
     X(vkCmdEndTransformFeedbackEXT);
@@ -221,6 +221,12 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     if (!dld.vkResetQueryPool) {
         Proc(dld.vkResetQueryPool, dld, "vkResetQueryPoolEXT", device);
     }
+
+    // Support for draw indirect with count is optional in Vulkan 1.2
+    if (!dld.vkCmdDrawIndirectCount) {
+        Proc(dld.vkCmdDrawIndirectCount, dld, "vkCmdDrawIndirectCountKHR", device);
+        Proc(dld.vkCmdDrawIndexedIndirectCount, dld, "vkCmdDrawIndexedIndirectCountKHR", device);
+    }
 #undef X
 }
 

src/video_core/vulkan_common/vulkan_wrapper.h

@@ -215,8 +215,8 @@ struct DeviceDispatch : InstanceDispatch {
     PFN_vkCmdDrawIndexed vkCmdDrawIndexed{};
     PFN_vkCmdDrawIndirect vkCmdDrawIndirect{};
     PFN_vkCmdDrawIndexedIndirect vkCmdDrawIndexedIndirect{};
-    PFN_vkCmdDrawIndirectCountKHR vkCmdDrawIndirectCountKHR{};
-    PFN_vkCmdDrawIndexedIndirectCountKHR vkCmdDrawIndexedIndirectCountKHR{};
+    PFN_vkCmdDrawIndirectCount vkCmdDrawIndirectCount{};
+    PFN_vkCmdDrawIndexedIndirectCount vkCmdDrawIndexedIndirectCount{};
     PFN_vkCmdEndDebugUtilsLabelEXT vkCmdEndDebugUtilsLabelEXT{};
     PFN_vkCmdEndQuery vkCmdEndQuery{};
     PFN_vkCmdEndRenderPass vkCmdEndRenderPass{};
@@ -1065,15 +1065,15 @@ public:
 
     void DrawIndirectCount(VkBuffer src_buffer, VkDeviceSize src_offset, VkBuffer count_buffer,
                            VkDeviceSize count_offset, u32 draw_count, u32 stride) const noexcept {
-        dld->vkCmdDrawIndirectCountKHR(handle, src_buffer, src_offset, count_buffer, count_offset,
-                                       draw_count, stride);
+        dld->vkCmdDrawIndirectCount(handle, src_buffer, src_offset, count_buffer, count_offset,
+                                    draw_count, stride);
     }
 
     void DrawIndexedIndirectCount(VkBuffer src_buffer, VkDeviceSize src_offset,
                                   VkBuffer count_buffer, VkDeviceSize count_offset, u32 draw_count,
                                   u32 stride) const noexcept {
-        dld->vkCmdDrawIndexedIndirectCountKHR(handle, src_buffer, src_offset, count_buffer,
-                                              count_offset, draw_count, stride);
+        dld->vkCmdDrawIndexedIndirectCount(handle, src_buffer, src_offset, count_buffer,
+                                           count_offset, draw_count, stride);
     }
 
     void ClearAttachments(Span<VkClearAttachment> attachments,

src/yuzu/Info.plist

@@ -34,6 +34,8 @@ SPDX-License-Identifier: GPL-2.0-or-later
     <string></string>
     <key>CSResourcesFileMapped</key>
     <true/>
+    <key>LSApplicationCategoryType</key>
+    <string>public.app-category.games</string>
     <key>LSRequiresCarbon</key>
     <true/>
     <key>NSHumanReadableCopyright</key>

src/yuzu/bootmanager.cpp

@@ -96,9 +96,9 @@ void EmuThread::run() {
             m_is_running.store(false);
             m_is_running.notify_all();
 
-            emit DebugModeEntered();
+            EmulationPaused(lk);
             Common::CondvarWait(m_should_run_cv, lk, stop_token, [&] { return m_should_run; });
-            emit DebugModeLeft();
+            EmulationResumed(lk);
         }
     }
 
@@ -111,6 +111,21 @@ void EmuThread::run() {
 #endif
 }
 
+// Unlock while emitting signals so that the main thread can
+// continue pumping events.
+
+void EmuThread::EmulationPaused(std::unique_lock<std::mutex>& lk) {
+    lk.unlock();
+    emit DebugModeEntered();
+    lk.lock();
+}
+
+void EmuThread::EmulationResumed(std::unique_lock<std::mutex>& lk) {
+    lk.unlock();
+    emit DebugModeLeft();
+    lk.lock();
+}
+
 #ifdef HAS_OPENGL
 class OpenGLSharedContext : public Core::Frontend::GraphicsContext {
 public:

src/yuzu/bootmanager.h

@@ -91,6 +91,10 @@ public:
         m_stop_source.request_stop();
     }
 
+private:
+    void EmulationPaused(std::unique_lock<std::mutex>& lk);
+    void EmulationResumed(std::unique_lock<std::mutex>& lk);
+
 private:
     Core::System& m_system;
 

src/yuzu/configuration/config.cpp

@@ -690,6 +690,7 @@ void Config::ReadRendererValues() {
     qt_config->beginGroup(QStringLiteral("Renderer"));
 
     ReadGlobalSetting(Settings::values.renderer_backend);
+    ReadGlobalSetting(Settings::values.renderer_force_max_clock);
     ReadGlobalSetting(Settings::values.vulkan_device);
     ReadGlobalSetting(Settings::values.fullscreen_mode);
     ReadGlobalSetting(Settings::values.aspect_ratio);
@@ -1306,6 +1307,9 @@ void Config::SaveRendererValues() {
                  static_cast<u32>(Settings::values.renderer_backend.GetValue(global)),
                  static_cast<u32>(Settings::values.renderer_backend.GetDefault()),
                  Settings::values.renderer_backend.UsingGlobal());
+    WriteSetting(QString::fromStdString(Settings::values.renderer_force_max_clock.GetLabel()),
+                 static_cast<u32>(Settings::values.renderer_force_max_clock.GetValue(global)),
+                 static_cast<u32>(Settings::values.renderer_force_max_clock.GetDefault()));
     WriteGlobalSetting(Settings::values.vulkan_device);
     WriteSetting(QString::fromStdString(Settings::values.fullscreen_mode.GetLabel()),
                  static_cast<u32>(Settings::values.fullscreen_mode.GetValue(global)),

src/yuzu/configuration/configure_graphics.ui

@@ -364,6 +364,11 @@
                <string>1X (720p/1080p)</string>
               </property>
              </item>
+             <item>
+              <property name="text">
+                  <string>1.5X (1080p/1620p) [EXPERIMENTAL]</string>
+              </property>
+             </item>
              <item>
               <property name="text">
                <string>2X (1440p/2160p)</string>
@@ -389,6 +394,16 @@
                <string>6X (4320p/6480p)</string>
               </property>
              </item>
+             <item>
+              <property name="text">
+               <string>7X (5040p/7560p)</string>
+              </property>
+             </item>
+             <item>
+              <property name="text">
+               <string>8X (5760p/8640p)</string>
+              </property>
+             </item>
             </widget>
            </item>
           </layout>

src/yuzu/configuration/configure_graphics_advanced.cpp

@@ -22,9 +22,11 @@ ConfigureGraphicsAdvanced::~ConfigureGraphicsAdvanced() = default;
 void ConfigureGraphicsAdvanced::SetConfiguration() {
     const bool runtime_lock = !system.IsPoweredOn();
     ui->use_vsync->setEnabled(runtime_lock);
+    ui->renderer_force_max_clock->setEnabled(runtime_lock);
     ui->use_asynchronous_shaders->setEnabled(runtime_lock);
     ui->anisotropic_filtering_combobox->setEnabled(runtime_lock);
 
+    ui->renderer_force_max_clock->setChecked(Settings::values.renderer_force_max_clock.GetValue());
     ui->use_vsync->setChecked(Settings::values.use_vsync.GetValue());
     ui->use_asynchronous_shaders->setChecked(Settings::values.use_asynchronous_shaders.GetValue());
     ui->use_fast_gpu_time->setChecked(Settings::values.use_fast_gpu_time.GetValue());
@@ -43,6 +45,8 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
                                                &Settings::values.max_anisotropy);
         ConfigurationShared::SetHighlight(ui->label_gpu_accuracy,
                                           !Settings::values.gpu_accuracy.UsingGlobal());
+        ConfigurationShared::SetHighlight(ui->renderer_force_max_clock,
+                                          !Settings::values.renderer_force_max_clock.UsingGlobal());
         ConfigurationShared::SetHighlight(ui->af_label,
                                           !Settings::values.max_anisotropy.UsingGlobal());
     }
@@ -50,6 +54,9 @@ void ConfigureGraphicsAdvanced::SetConfiguration() {
 
 void ConfigureGraphicsAdvanced::ApplyConfiguration() {
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.gpu_accuracy, ui->gpu_accuracy);
+    ConfigurationShared::ApplyPerGameSetting(&Settings::values.renderer_force_max_clock,
+                                             ui->renderer_force_max_clock,
+                                             renderer_force_max_clock);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.max_anisotropy,
                                              ui->anisotropic_filtering_combobox);
     ConfigurationShared::ApplyPerGameSetting(&Settings::values.use_vsync, ui->use_vsync, use_vsync);
@@ -81,6 +88,8 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
     // Disable if not global (only happens during game)
     if (Settings::IsConfiguringGlobal()) {
         ui->gpu_accuracy->setEnabled(Settings::values.gpu_accuracy.UsingGlobal());
+        ui->renderer_force_max_clock->setEnabled(
+            Settings::values.renderer_force_max_clock.UsingGlobal());
         ui->use_vsync->setEnabled(Settings::values.use_vsync.UsingGlobal());
         ui->use_asynchronous_shaders->setEnabled(
             Settings::values.use_asynchronous_shaders.UsingGlobal());
@@ -95,6 +104,9 @@ void ConfigureGraphicsAdvanced::SetupPerGameUI() {
         return;
     }
 
+    ConfigurationShared::SetColoredTristate(ui->renderer_force_max_clock,
+                                            Settings::values.renderer_force_max_clock,
+                                            renderer_force_max_clock);
     ConfigurationShared::SetColoredTristate(ui->use_vsync, Settings::values.use_vsync, use_vsync);
     ConfigurationShared::SetColoredTristate(ui->use_asynchronous_shaders,
                                             Settings::values.use_asynchronous_shaders,

src/yuzu/configuration/configure_graphics_advanced.h

@@ -36,6 +36,7 @@ private:
 
     std::unique_ptr<Ui::ConfigureGraphicsAdvanced> ui;
 
+    ConfigurationShared::CheckState renderer_force_max_clock;
     ConfigurationShared::CheckState use_vsync;
     ConfigurationShared::CheckState use_asynchronous_shaders;
     ConfigurationShared::CheckState use_fast_gpu_time;

src/yuzu/configuration/configure_graphics_advanced.ui

@@ -69,6 +69,16 @@
           </layout>
          </widget>
         </item>
+        <item>
+         <widget class="QCheckBox" name="renderer_force_max_clock">
+          <property name="toolTip">
+           <string>Runs work in the background while waiting for graphics commands to keep the GPU from lowering its clock speed.</string>
+          </property>
+          <property name="text">
+           <string>Force maximum clocks (Vulkan only)</string>
+          </property>
+         </widget>
+        </item>
         <item>
          <widget class="QCheckBox" name="use_vsync">
           <property name="toolTip">

src/yuzu/configuration/configure_input_player.cpp

@@ -1466,6 +1466,12 @@ void ConfigureInputPlayer::mousePressEvent(QMouseEvent* event) {
     input_subsystem->GetMouse()->PressButton(0, 0, 0, 0, button);
 }
 
+void ConfigureInputPlayer::wheelEvent(QWheelEvent* event) {
+    const int x = event->angleDelta().x();
+    const int y = event->angleDelta().y();
+    input_subsystem->GetMouse()->MouseWheelChange(x, y);
+}
+
 void ConfigureInputPlayer::keyPressEvent(QKeyEvent* event) {
     if (!input_setter || !event) {
         return;

src/yuzu/configuration/configure_input_player.h

@@ -116,6 +116,9 @@ private:
     /// Handle mouse button press events.
     void mousePressEvent(QMouseEvent* event) override;
 
+    /// Handle mouse wheel move events.
+    void wheelEvent(QWheelEvent* event) override;
+
     /// Handle key press events.
     void keyPressEvent(QKeyEvent* event) override;
 

src/yuzu/main.cpp

@@ -1839,9 +1839,11 @@ void GMainWindow::OnEmulationStopTimeExpired() {
 
 void GMainWindow::OnEmulationStopped() {
     shutdown_timer.stop();
-    emu_thread->disconnect();
-    emu_thread->wait();
-    emu_thread = nullptr;
+    if (emu_thread) {
+        emu_thread->disconnect();
+        emu_thread->wait();
+        emu_thread.reset();
+    }
 
     if (shutdown_dialog) {
         shutdown_dialog->deleteLater();
@@ -3029,6 +3031,8 @@ void GMainWindow::OnStopGame() {
 
     if (OnShutdownBegin()) {
         OnShutdownBeginDialog();
+    } else {
+        OnEmulationStopped();
     }
 }
 
@@ -3726,15 +3730,36 @@ void GMainWindow::UpdateWindowTitle(std::string_view title_name, std::string_vie
     }
 }
 
+std::string GMainWindow::CreateTASFramesString(
+    std::array<size_t, InputCommon::TasInput::PLAYER_NUMBER> frames) const {
+    std::string string = "";
+    size_t maxPlayerIndex = 0;
+    for (size_t i = 0; i < frames.size(); i++) {
+        if (frames[i] != 0) {
+            if (maxPlayerIndex != 0)
+                string += ", ";
+            while (maxPlayerIndex++ != i)
+                string += "0, ";
+            string += std::to_string(frames[i]);
+        }
+    }
+    return string;
+}
+
 QString GMainWindow::GetTasStateDescription() const {
     auto [tas_status, current_tas_frame, total_tas_frames] = input_subsystem->GetTas()->GetStatus();
+    std::string tas_frames_string = CreateTASFramesString(total_tas_frames);
     switch (tas_status) {
     case InputCommon::TasInput::TasState::Running:
-        return tr("TAS state: Running %1/%2").arg(current_tas_frame).arg(total_tas_frames);
+        return tr("TAS state: Running %1/%2")
+            .arg(current_tas_frame)
+            .arg(QString::fromStdString(tas_frames_string));
     case InputCommon::TasInput::TasState::Recording:
-        return tr("TAS state: Recording %1").arg(total_tas_frames);
+        return tr("TAS state: Recording %1").arg(total_tas_frames[0]);
     case InputCommon::TasInput::TasState::Stopped:
-        return tr("TAS state: Idle %1/%2").arg(current_tas_frame).arg(total_tas_frames);
+        return tr("TAS state: Idle %1/%2")
+            .arg(current_tas_frame)
+            .arg(QString::fromStdString(tas_frames_string));
     default:
         return tr("TAS State: Invalid");
     }

src/yuzu/main.h

@@ -12,6 +12,7 @@
 
 #include "common/announce_multiplayer_room.h"
 #include "common/common_types.h"
+#include "input_common/drivers/tas_input.h"
 #include "yuzu/compatibility_list.h"
 #include "yuzu/hotkeys.h"
 
@@ -266,6 +267,9 @@ private:
     void changeEvent(QEvent* event) override;
     void closeEvent(QCloseEvent* event) override;
 
+    std::string CreateTASFramesString(
+        std::array<size_t, InputCommon::TasInput::PLAYER_NUMBER> frames) const;
+
 #ifdef __unix__
     void SetupSigInterrupts();
     static void HandleSigInterrupt(int);

src/yuzu_cmd/config.cpp

@@ -296,6 +296,7 @@ void Config::ReadValues() {
 
     // Renderer
     ReadSetting("Renderer", Settings::values.renderer_backend);
+    ReadSetting("Renderer", Settings::values.renderer_force_max_clock);
     ReadSetting("Renderer", Settings::values.renderer_debug);
     ReadSetting("Renderer", Settings::values.renderer_shader_feedback);
     ReadSetting("Renderer", Settings::values.enable_nsight_aftermath);