AAaaa

(cherry picked from commit 9c313fb787)

.github/PULL_REQUEST_TEMPLATE.md

@@ -1,5 +0,0 @@
-- [ ] I have read and followed the [Contribution Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/CONTRIBUTING.md#code-contributions).
-- [ ] I have read and followed the [AI Policy](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/AI.md)
-- [ ] I have read and followed the [Coding Guidelines](https://git.eden-emu.dev/eden-emu/eden/src/branch/master/docs/policies/Coding.md) to the best of my ability.
-
--------------------

src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt

@@ -594,7 +594,7 @@ abstract class SettingsItem(
                     IntSetting.ANDROID_PIPELINE_WORKERS,
                     titleId = R.string.pipeline_worker_cores,
                     descriptionId = R.string.pipeline_worker_cores_description,
-                    min = 4,
+                    min = 1,
                     max = 8,
                     units = "cores"
                 )

src/android/app/src/main/jni/android_settings.h

@@ -147,7 +147,7 @@ namespace AndroidSettings {
                                                       &show_performance_overlay};
 
 
-        Settings::Setting<s32> pipeline_worker_count{linkage, 4, "pipeline_worker_count",
+        Settings::Setting<s32> pipeline_worker_count{linkage, 2, "pipeline_worker_count",
                                  Settings::Category::Android,
                                  Settings::Specialization::Default,
                                  true,

src/android/app/src/main/res/values-ar/strings.xml

@@ -1000,6 +1000,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">سريع</string>
+    <string name="renderer_accuracy_medium">متوازن</string>
     <string name="renderer_accuracy_high">دقيق</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-cs/strings.xml

@@ -707,6 +707,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Rychlý</string>
+    <string name="renderer_accuracy_medium">Vyvážený</string>
     <string name="renderer_accuracy_high">Přesný</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-de/strings.xml

@@ -927,6 +927,7 @@ Wirklich fortfahren?</string>
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Schnell</string>
+    <string name="renderer_accuracy_medium">Ausgeglichen</string>
     <string name="renderer_accuracy_high">Genau</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-es/strings.xml

@@ -993,6 +993,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Rápido</string>
+    <string name="renderer_accuracy_medium">Equilibrado</string>
     <string name="renderer_accuracy_high">Preciso</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-fr/strings.xml

@@ -939,6 +939,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Rapide</string>
+    <string name="renderer_accuracy_medium">Moyen</string>
     <string name="renderer_accuracy_high">Précis</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-pl/strings.xml

@@ -893,6 +893,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Szybkie</string>
+    <string name="renderer_accuracy_medium">Zrównoważony</string>
     <string name="renderer_accuracy_high">Dokładny</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-pt-rBR/strings.xml

@@ -840,6 +840,7 @@
 
     <string name="renderer_none">Nenhum</string>
 
+    <string name="renderer_accuracy_medium">Média</string>
     <string name="renderer_accuracy_high">Alta</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-ru/strings.xml

@@ -983,6 +983,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Быстрый</string>
+    <string name="renderer_accuracy_medium">Сбалансированный</string>
     <string name="renderer_accuracy_high">Точный</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-uk/strings.xml

@@ -986,6 +986,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Швидко</string>
+    <string name="renderer_accuracy_medium">Збалансовано</string>
     <string name="renderer_accuracy_high">Точно</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-zh-rCN/strings.xml

@@ -990,6 +990,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">快速</string>
+    <string name="renderer_accuracy_medium">均衡</string>
     <string name="renderer_accuracy_high">精确</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values-zh-rTW/strings.xml

@@ -843,6 +843,7 @@
 
     <string name="renderer_none">無</string>
 
+    <string name="renderer_accuracy_medium">平衡</string>
     <string name="renderer_accuracy_high">準確</string>
 
     <!-- DMA Accuracy -->

src/android/app/src/main/res/values/arrays.xml

@@ -103,12 +103,14 @@
 
     <string-array name="rendererAccuracyNames">
         <item>@string/renderer_accuracy_low</item>
+        <item>@string/renderer_accuracy_medium</item>
         <item>@string/renderer_accuracy_high</item>
     </string-array>
 
     <integer-array name="rendererAccuracyValues">
         <item>0</item>
         <item>1</item>
+        <item>2</item>
     </integer-array>
 
     <!-- VRAM USAGE MODE CHOICES -->

src/android/app/src/main/res/values/strings.xml

@@ -479,7 +479,7 @@
     <string name="advanced">Advanced</string>
 
     <string name="renderer_accuracy">GPU Mode</string>
-    <string name="renderer_accuracy_description">Controls the GPU emulation mode. Most games render fine with Fast, but Accurate is still required for some. Particles tend to only render correctly with Accurate mode.</string>
+    <string name="renderer_accuracy_description">Controls the GPU emulation mode. Most games render fine with Fast or Balanced modes, but Accurate is still required for some. Particles tend to only render correctly with Accurate mode.</string>
     <string name="dma_accuracy">DMA Accuracy</string>
     <string name="dma_accuracy_description">Controls the DMA precision accuracy. Safe precision can fix issues in some games, but it can also impact performance in some cases. If unsure, leave this on Default.</string>
     <string name="gpu_fence_behavior">GPU Fence Behavior</string>
@@ -1039,6 +1039,7 @@
 
     <!-- Renderer Accuracy -->
     <string name="renderer_accuracy_low">Fast</string>
+    <string name="renderer_accuracy_medium">Balanced</string>
     <string name="renderer_accuracy_high">Accurate</string>
 
     <!-- DMA Accuracy -->

src/common/fs/path_util.cpp

@@ -476,29 +476,6 @@ std::string SanitizePath(std::string_view path_, DirectorySeparator directory_se
     path.erase(std::unique(start, path.end(),
                            [type2](char c1, char c2) { return c1 == type2 && c2 == type2; }),
                path.end());
-
-    const bool absolute = !path.empty() && path[0] == type2;
-    std::vector<std::string_view> parts;
-
-    for (const auto part : SplitPathComponents(path))
-    {
-        if (part.empty() || part == ".")
-            continue;
-        if (part == ".." && !parts.empty() && parts.back() != "..")
-            parts.pop_back();
-        else if (part != "..") parts.push_back(part);
-    }
-
-    std::string resolved = absolute ? std::string(1, type2) : std::string{};
-    for (std::size_t i = 0; i < parts.size(); ++i)
-    {
-        if (i != 0)
-            resolved += type2;
-        resolved.append(parts[i].data(), parts[i].size());
-    }
-
-    path = std::move(resolved);
-
     return std::string(RemoveTrailingSlash(path));
 }
 

src/core/arm/nce/arm_nce.cpp

@@ -157,6 +157,8 @@ bool ArmNce::HandleGuestAlignmentFault(GuestContext* guest_ctx, void* raw_info,
     return HandleFailedGuestFault(guest_ctx, raw_info, raw_context);
 }
 
+constexpr size_t NCE_WRITE_FAULT_CLUSTER_PAGES = 4;
+
 bool ArmNce::HandleGuestAccessFault(GuestContext* guest_ctx, void* raw_info, void* raw_context) {
     auto* info = static_cast<siginfo_t*>(raw_info);
 
@@ -165,7 +167,7 @@ bool ArmNce::HandleGuestAccessFault(GuestContext* guest_ctx, void* raw_info, voi
     const Common::ProcessAddress addr =
         (reinterpret_cast<u64>(info->si_addr) & ~Memory::YUZU_PAGEMASK);
     auto& memory = guest_ctx->parent->m_running_thread->GetOwnerProcess()->GetMemory();
-    if (memory.InvalidateNCE(addr, Memory::YUZU_PAGESIZE)) {
+    if (memory.InvalidateNCE(addr, Memory::YUZU_PAGESIZE * NCE_WRITE_FAULT_CLUSTER_PAGES)) {
         // We handled the access successfully and are returning to guest code.
         return true;
     }

src/core/arm/nce/patcher.cpp

@@ -145,14 +145,7 @@ bool Patcher::PatchText(std::span<const u8> program_image, const Kernel::CodeSet
 
         // MRS Xn, CNTFRQ_EL0
         if (auto mrs = MRS{inst}; mrs.Verify() && mrs.GetSystemReg() == CntfrqEl0) {
-            bool pre_buffer = false;
-            auto ret = AddRelocations(pre_buffer);
-            if (pre_buffer) {
-                WriteCntfrqHandler(ret, oaknut::XReg{static_cast<int>(mrs.GetRt())}, c_pre);
-            } else {
-                WriteCntfrqHandler(ret, oaknut::XReg{static_cast<int>(mrs.GetRt())}, c);
-            }
-            continue;
+            UNREACHABLE();
         }
 
         // MSR TPIDR_EL0, Xn
@@ -584,16 +577,6 @@ void Patcher::WriteMsrHandler(ModuleDestLabel module_dest, oaknut::XReg src_reg,
         this->BranchToModule(module_dest);
 }
 
-void Patcher::WriteCntfrqHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg, oaknut::VectorCodeGenerator& cg) {
-    cg.MOV(dest_reg, Common::WallClock::CNTFRQ);
-
-    // Jump back to the instruction after the emulated MRS.
-    if (&cg == &c_pre)
-        this->BranchToModulePre(module_dest);
-    else
-        this->BranchToModule(module_dest);
-}
-
 void Patcher::WriteCntpctHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg, oaknut::VectorCodeGenerator& cg) {
 #if defined(HAS_NCE)
     static Common::WallClock clock(false, 1);

src/core/arm/nce/patcher.h

@@ -80,7 +80,6 @@ private:
     void WriteSvcTrampoline(ModuleDestLabel module_dest, u32 svc_id, oaknut::VectorCodeGenerator& code, oaknut::Label& save_ctx, oaknut::Label& load_ctx);
     void WriteMrsHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg, oaknut::SystemReg src_reg, oaknut::VectorCodeGenerator& code);
     void WriteMsrHandler(ModuleDestLabel module_dest, oaknut::XReg src_reg, oaknut::VectorCodeGenerator& code);
-    void WriteCntfrqHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg, oaknut::VectorCodeGenerator& code);
     void WriteCntpctHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg, oaknut::VectorCodeGenerator& code);
 
     // Convenience wrappers using default code generator
@@ -91,7 +90,6 @@ private:
     void WriteSvcTrampoline(ModuleDestLabel module_dest, u32 svc_id) { WriteSvcTrampoline(module_dest, svc_id, c, m_save_context, m_load_context); }
     void WriteMrsHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg, oaknut::SystemReg src_reg) { WriteMrsHandler(module_dest, dest_reg, src_reg, c); }
     void WriteMsrHandler(ModuleDestLabel module_dest, oaknut::XReg src_reg) { WriteMsrHandler(module_dest, src_reg, c); }
-    void WriteCntfrqHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg) { WriteCntfrqHandler(module_dest, dest_reg, c); }
     void WriteCntpctHandler(ModuleDestLabel module_dest, oaknut::XReg dest_reg) { WriteCntpctHandler(module_dest, dest_reg, c); }
 
 private:

src/core/device_memory_manager.h

@@ -126,6 +126,10 @@ public:
     // New batch API to update multiple ranges with a single lock acquisition.
     void UpdatePagesCachedBatch(std::span<const std::pair<DAddr, size_t>> ranges, s32 delta);
 
+    void UpdateTexturePagesCount(DAddr addr, size_t size, s32 delta);
+
+    [[nodiscard]] bool IsRegionTextureCached(DAddr addr, size_t size) const noexcept;
+
 private:
     struct TranslationEntry {
         DAddr guest_page{};
@@ -234,6 +238,7 @@ private:
         (1ULL << (device_virtual_bits - page_bits)) / subentries;
     using CachedPages = std::array<CounterEntry, num_counter_entries>;
     std::unique_ptr<CachedPages> cached_pages;
+    std::unique_ptr<CachedPages> texture_cached_pages;
     Common::RangeMutex counter_guard;
     std::mutex mapping_guard;
 

src/core/device_memory_manager.inc

@@ -177,6 +177,7 @@ DeviceMemoryManager<Traits>::DeviceMemoryManager(const DeviceMemory& device_memo
 {
     impl = std::make_unique<DeviceMemoryManagerAllocator<Traits>>();
     cached_pages = std::make_unique<CachedPages>();
+    texture_cached_pages = std::make_unique<CachedPages>();
 
     const size_t total_virtual = device_as_size >> Memory::YUZU_PAGEBITS;
     for (size_t i = 0; i < total_virtual; i++) {
@@ -625,6 +626,28 @@ void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size
     UpdatePagesCachedCountNoLock(addr, size, delta);
 }
 
+template <typename Traits>
+void DeviceMemoryManager<Traits>::UpdateTexturePagesCount(DAddr addr, size_t size, s32 delta) {
+    Common::ScopedRangeLock lk(counter_guard, addr, size);
+    const size_t page_end = Common::DivCeil(addr + size, Memory::YUZU_PAGESIZE);
+    for (size_t page = addr >> Memory::YUZU_PAGEBITS; page != page_end; ++page) {
+        CounterAtomicType& count = texture_cached_pages->at(page >> subentries_shift).Count(page);
+        count.fetch_add(static_cast<CounterType>(delta), std::memory_order_release);
+    }
+}
+
+template <typename Traits>
+bool DeviceMemoryManager<Traits>::IsRegionTextureCached(DAddr addr, size_t size) const noexcept {
+    const size_t page_end = Common::DivCeil(addr + size, Memory::YUZU_PAGESIZE);
+    for (size_t page = addr >> Memory::YUZU_PAGEBITS; page != page_end; ++page) {
+        if (texture_cached_pages->at(page >> subentries_shift).Count(page).load(
+                std::memory_order_acquire) != 0) {
+            return true;
+        }
+    }
+    return false;
+}
+
 template <typename Traits>
 void DeviceMemoryManager<Traits>::UpdatePagesCachedBatch(std::span<const std::pair<DAddr, size_t>> ranges, s32 delta) {
     if (ranges.empty()) {

src/core/file_sys/registered_cache.cpp

@@ -984,22 +984,6 @@ bool RegisteredCache::RemoveExistingEntry(u64 title_id) const {
     return removed_data;
 }
 
-bool RegisteredCache::Delete(const NcaID& id) const {
-    const auto path = GetRelativePathFromNcaID(id, false, true, false);
-
-    const bool is_file = dir->GetFileRelative(path) != nullptr;
-    const bool is_dir = dir->GetDirectoryRelative(path) != nullptr;
-
-    if (is_file) {
-        return dir->DeleteFile(path);
-    }
-    if (is_dir) {
-        return dir->DeleteSubdirectoryRecursive(path);
-    }
-
-    return true;
-}
-
 InstallResult RegisteredCache::RawInstallNCA(const NCA& nca, const VfsCopyFunction& copy,
                                              bool overwrite_if_exists,
                                              std::optional<NcaID> override_id) {

src/core/file_sys/registered_cache.h

@@ -188,7 +188,6 @@ public:
 
     // Removes an existing entry based on title id
     bool RemoveExistingEntry(u64 title_id) const;
-    bool Delete(const NcaID& id) const;
 
 private:
     template <typename T>

src/core/file_sys/vfs/vfs_real.cpp

@@ -35,16 +35,6 @@ namespace {
 
 constexpr size_t MaxOpenFiles = 8192;
 
-bool IsWithinRoot(std::string_view root, std::string_view full_path) {
-    if (root.empty())
-        return true;
-
-    if (full_path.size() < root.size() || full_path.substr(0, root.size()) != root)
-        return false;
-
-    return full_path.size() == root.size() || full_path[root.size()] == '/' || full_path[root.size()] == '\\';
-}
-
 constexpr FS::FileAccessMode ModeFlagsToFileAccessMode(OpenMode mode) {
     switch (mode) {
     case OpenMode::Read:
@@ -413,8 +403,7 @@ RealVfsDirectory::~RealVfsDirectory() = default;
 
 VirtualFile RealVfsDirectory::GetFileRelative(std::string_view relative_path) const {
     const auto full_path = FS::SanitizePath(path + '/' + std::string(relative_path));
-    if (!FS::Exists(full_path) || FS::IsDir(full_path)
-        || !IsWithinRoot(FS::SanitizePath(path), full_path)) {
+    if (!FS::Exists(full_path) || FS::IsDir(full_path)) {
         return nullptr;
     }
     return base.OpenFile(full_path, perms);
@@ -422,8 +411,7 @@ VirtualFile RealVfsDirectory::GetFileRelative(std::string_view relative_path) co
 
 VirtualDir RealVfsDirectory::GetDirectoryRelative(std::string_view relative_path) const {
     const auto full_path = FS::SanitizePath(path + '/' + std::string(relative_path));
-    if (!FS::Exists(full_path) || !FS::IsDir(full_path)
-        || !IsWithinRoot(FS::SanitizePath(path), full_path)) {
+    if (!FS::Exists(full_path) || !FS::IsDir(full_path)) {
         return nullptr;
     }
     return base.OpenDirectory(full_path, perms);
@@ -439,7 +427,7 @@ VirtualDir RealVfsDirectory::GetSubdirectory(std::string_view name) const {
 
 VirtualFile RealVfsDirectory::CreateFileRelative(std::string_view relative_path) {
     const auto full_path = FS::SanitizePath(path + '/' + std::string(relative_path));
-    if (!FS::CreateParentDirs(full_path) || !IsWithinRoot(FS::SanitizePath(path), full_path)) {
+    if (!FS::CreateParentDirs(full_path)) {
         return nullptr;
     }
     return base.CreateFile(full_path, perms);
@@ -452,7 +440,7 @@ VirtualDir RealVfsDirectory::CreateDirectoryRelative(std::string_view relative_p
 
 bool RealVfsDirectory::DeleteSubdirectoryRecursive(std::string_view name) {
     const auto full_path = FS::SanitizePath(this->path + '/' + std::string(name));
-    return FS::RemoveDirRecursively(full_path);
+    return base.DeleteDirectory(full_path);
 }
 
 std::vector<VirtualFile> RealVfsDirectory::GetFiles() const {
@@ -518,7 +506,7 @@ VirtualFile RealVfsDirectory::CreateFile(std::string_view name) {
 
 bool RealVfsDirectory::DeleteSubdirectory(std::string_view name) {
     const std::string subdir_path = (path + '/').append(name);
-    return FS::RemoveDir(subdir_path);
+    return base.DeleteDirectory(subdir_path);
 }
 
 bool RealVfsDirectory::DeleteFile(std::string_view name) {

src/core/hle/service/am/service/self_controller.cpp

@@ -57,7 +57,7 @@ ISelfController::ISelfController(Core::System& system_, std::shared_ptr<Applet>
         {64, nullptr, "SetInputDetectionSourceSet"},
         {65, D<&ISelfController::ReportUserIsActive>, "ReportUserIsActive"},
         {66, nullptr, "GetCurrentIlluminance"},
-        {67, D<&ISelfController::IsIlluminanceAvailable>, "IsIlluminanceAvailable"},
+        {67, nullptr, "IsIlluminanceAvailable"},
         {68, D<&ISelfController::SetAutoSleepDisabled>, "SetAutoSleepDisabled"},
         {69, D<&ISelfController::IsAutoSleepDisabled>, "IsAutoSleepDisabled"},
         {70, nullptr, "ReportMultimediaError"},
@@ -347,12 +347,6 @@ Result ISelfController::IsAutoSleepDisabled(Out<bool> out_is_auto_sleep_disabled
     R_SUCCEED();
 }
 
-Result ISelfController::IsIlluminanceAvailable(Out<bool> out_is_illuminance_available) {
-    LOG_WARNING(Service_AM, "(stubbed)");
-    *out_is_illuminance_available = false;
-    R_SUCCEED();
-}
-
 Result ISelfController::SetInputDetectionPolicy(InputDetectionPolicy input_detection_policy) {
     LOG_WARNING(Service_AM, "(STUBBED) called");
     R_SUCCEED();

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

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2024 yuzu Emulator Project
@@ -60,7 +60,6 @@ private:
     Result ReportUserIsActive();
     Result SetAutoSleepDisabled(bool is_auto_sleep_disabled);
     Result IsAutoSleepDisabled(Out<bool> out_is_auto_sleep_disabled);
-    Result IsIlluminanceAvailable(Out<bool> out_is_illuminance_available);
     Result SetInputDetectionPolicy(InputDetectionPolicy input_detection_policy);
     Result GetAccumulatedSuspendedTickValue(Out<u64> out_accumulated_suspended_tick_value);
     Result GetAccumulatedSuspendedTickChangedEvent(OutCopyHandle<Kernel::KReadableEvent> out_event);

src/core/hle/service/filesystem/filesystem.cpp

@@ -43,16 +43,6 @@ static FileSys::VirtualDir GetDirectoryRelativeWrapped(FileSys::VirtualDir base,
     return base->GetDirectoryRelative(dir_name);
 }
 
-static std::string_view GetGuestParentPath(std::string_view path) {
-    const auto name_index = path.find_last_of("\\/");
-    return name_index == std::string_view::npos ? std::string_view{} : path.substr(0, name_index);
-}
-
-static std::string_view GetGuestFilename(std::string_view path) {
-    const auto name_index = path.find_last_of("\\/");
-    return name_index == std::string_view::npos ? path : path.substr(name_index + 1);
-}
-
 VfsDirectoryServiceWrapper::VfsDirectoryServiceWrapper(FileSys::VirtualDir backing_)
     : backing(std::move(backing_)) {}
 
@@ -93,12 +83,11 @@ Result VfsDirectoryServiceWrapper::DeleteFile(const std::string& path_) const {
         return ResultSuccess;
     }
 
-    const auto filename = GetGuestFilename(path);
-    auto dir = GetDirectoryRelativeWrapped(backing, GetGuestParentPath(path));
-    if (filename.empty() || dir == nullptr || dir->GetFile(filename) == nullptr) {
+    auto dir = GetDirectoryRelativeWrapped(backing, Common::FS::GetParentPath(path));
+    if (dir == nullptr || dir->GetFile(Common::FS::GetFilename(path)) == nullptr) {
         return FileSys::ResultPathNotFound;
     }
-    if (!dir->DeleteFile(filename)) {
+    if (!dir->DeleteFile(Common::FS::GetFilename(path))) {
         // TODO(DarkLordZach): Find a better error code for this
         return ResultUnknown;
     }
@@ -108,9 +97,7 @@ Result VfsDirectoryServiceWrapper::DeleteFile(const std::string& path_) const {
 
 Result VfsDirectoryServiceWrapper::CreateDirectory(const std::string& path_) const {
     std::string path(Common::FS::SanitizePath(path_));
-    if (GetDirectoryRelativeWrapped(backing, path) != nullptr) {
-        return FileSys::ResultPathAlreadyExists;
-    }
+
     // NOTE: This is inaccurate behavior. CreateDirectory is not recursive.
     // CreateDirectory should return PathNotFound if the parent directory does not exist.
     // This is here temporarily in order to have UMM "work" in the meantime.
@@ -130,19 +117,8 @@ Result VfsDirectoryServiceWrapper::CreateDirectory(const std::string& path_) con
 
 Result VfsDirectoryServiceWrapper::DeleteDirectory(const std::string& path_) const {
     std::string path(Common::FS::SanitizePath(path_));
-    const auto dirname = GetGuestFilename(path);
-    auto dir = GetDirectoryRelativeWrapped(backing, GetGuestParentPath(path));
-    FileSys::VirtualDir target{};
-    if (!dirname.empty() && dir != nullptr) {
-        target = dir->GetSubdirectory(dirname);
-    }
-    if (target == nullptr) {
-        return FileSys::ResultPathNotFound;
-    }
-    if (!target->GetFiles().empty() || !target->GetSubdirectories().empty()) {
-        return ResultUnknown;
-    }
-    if (!dir->DeleteSubdirectory(dirname)) {
+    auto dir = GetDirectoryRelativeWrapped(backing, Common::FS::GetParentPath(path));
+    if (!dir->DeleteSubdirectory(Common::FS::GetFilename(path))) {
         // TODO(DarkLordZach): Find a better error code for this
         return ResultUnknown;
     }
@@ -151,12 +127,8 @@ Result VfsDirectoryServiceWrapper::DeleteDirectory(const std::string& path_) con
 
 Result VfsDirectoryServiceWrapper::DeleteDirectoryRecursively(const std::string& path_) const {
     std::string path(Common::FS::SanitizePath(path_));
-    const auto dirname = GetGuestFilename(path);
-    auto dir = GetDirectoryRelativeWrapped(backing, GetGuestParentPath(path));
-    if (dirname.empty() || dir == nullptr || dir->GetSubdirectory(dirname) == nullptr) {
-        return FileSys::ResultPathNotFound;
-    }
-    if (!dir->DeleteSubdirectoryRecursive(dirname)) {
+    auto dir = GetDirectoryRelativeWrapped(backing, Common::FS::GetParentPath(path));
+    if (!dir->DeleteSubdirectoryRecursive(Common::FS::GetFilename(path))) {
         // TODO(DarkLordZach): Find a better error code for this
         return ResultUnknown;
     }
@@ -165,13 +137,9 @@ Result VfsDirectoryServiceWrapper::DeleteDirectoryRecursively(const std::string&
 
 Result VfsDirectoryServiceWrapper::CleanDirectoryRecursively(const std::string& path) const {
     const std::string sanitized_path(Common::FS::SanitizePath(path));
-    const auto dirname = GetGuestFilename(sanitized_path);
-    auto dir = GetDirectoryRelativeWrapped(backing, GetGuestParentPath(sanitized_path));
+    auto dir = GetDirectoryRelativeWrapped(backing, Common::FS::GetParentPath(sanitized_path));
 
-    if (dirname.empty() || dir == nullptr || dir->GetSubdirectory(dirname) == nullptr) {
-        return FileSys::ResultPathNotFound;
-    }
-    if (!dir->CleanSubdirectoryRecursive(dirname)) {
+    if (!dir->CleanSubdirectoryRecursive(Common::FS::GetFilename(sanitized_path))) {
         // TODO(DarkLordZach): Find a better error code for this
         return ResultUnknown;
     }

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

@@ -237,7 +237,7 @@ IHidServer::IHidServer(Core::System& system_, std::shared_ptr<ResourceManager> r
         {3013, nullptr, "SetDebugPadGenericPadMap"}, //21.0.0+
         {3014, nullptr, "GetDebugPadKeyboardMap"}, //21.0.0+
         {3015, nullptr, "SetDebugPadKeyboardMap"}, //21.0.0+
-        {3150, C<&IHidServer::SetMouseLibraryVersion>, "SetMouseLibraryVersion"}, //21.0.0+
+        {3150, nullptr, "SetMouseLibraryVersion"}, //21.0.0+
         // What? -- {12010, nullptr, "SetButtonConfigLeft"},
     };
     // clang-format on
@@ -1471,12 +1471,6 @@ Result IHidServer::SetTouchScreenResolution(u32 width, u32 height,
     R_SUCCEED();
 }
 
-Result IHidServer::SetMouseLibraryVersion(ClientAppletResourceUserId aruid) {
-    LOG_INFO(Service_HID, "(STUBBED) called, applet_resource_user_id={}", aruid.pid);
-
-    R_SUCCEED();
-}
-
 std::shared_ptr<ResourceManager> IHidServer::GetResourceManager() {
     resource_manager->Initialize();
     return resource_manager;

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

@@ -1,4 +1,4 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-FileCopyrightText: Copyright 2025 Eden Emulator Project
 // SPDX-License-Identifier: GPL-3.0-or-later
 
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
@@ -263,7 +263,6 @@ private:
     Result IsFirmwareUpdateNeededForNotification(Out<bool> out_is_firmware_update_needed,
                                                  s32 unknown, ClientAppletResourceUserId aruid);
     Result SetTouchScreenResolution(u32 width, u32 height, ClientAppletResourceUserId aruid);
-    Result SetMouseLibraryVersion(ClientAppletResourceUserId aruid);
 
     std::shared_ptr<ResourceManager> resource_manager;
     std::shared_ptr<HidFirmwareSettings> firmware_settings;

src/core/hle/service/ncm/ncm.cpp

@@ -1,19 +1,9 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <algorithm>
-#include <array>
 #include <memory>
-#include <vector>
 
-#include "core/core.h"
-#include "core/file_sys/registered_cache.h"
 #include "core/file_sys/romfs_factory.h"
-#include "core/hle/api_version.h"
-#include "core/hle/service/filesystem/filesystem.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/ncm/ncm.h"
 #include "core/hle/service/server_manager.h"
@@ -98,268 +88,6 @@ public:
     }
 };
 
-class IContentStorage final : public ServiceFramework<IContentStorage> {
-public:
-    explicit IContentStorage(Core::System& system_, FileSys::StorageId id)
-        : ServiceFramework{system_, "IContentStorage"}, storage{id} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {0, &IContentStorage::GeneratePlaceHolderId, "GeneratePlaceHolderId"},
-            {1, &IContentStorage::CreatePlaceHolder, "CreatePlaceHolder"},
-            {2, &IContentStorage::DeletePlaceHolder, "DeletePlaceHolder"},
-            {4, &IContentStorage::WritePlaceHolder, "WritePlaceHolder"},
-            {5, &IContentStorage::Register, "Register"},
-            {6, &IContentStorage::Delete, "Delete"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-
-private:
-    void GeneratePlaceHolderId(HLERequestContext& ctx) {
-        LOG_DEBUG(Service_NCM, "called");
-
-        IPC::ResponseBuilder rb{ctx, 6};
-        rb.Push(ResultSuccess);
-        rb.PushRaw(FileSys::PlaceholderCache::Generate());
-    }
-
-    void CreatePlaceHolder(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        [[maybe_unused]] FileSys::NcaID content_id{};
-        FileSys::NcaID placeholder_id{};
-        if constexpr (HLE::ApiVersion::HOS_VERSION_MAJOR >= 16) {
-            placeholder_id = rp.PopRaw<FileSys::NcaID>();
-            content_id = rp.PopRaw<FileSys::NcaID>();
-        } else {
-            content_id = rp.PopRaw<FileSys::NcaID>();
-            placeholder_id = rp.PopRaw<FileSys::NcaID>();
-        }
-        const auto size = rp.Pop<s64>();
-
-        auto* const placeholder_cache =
-            system.GetFileSystemController().GetPlaceholderCacheForStorage(storage);
-        const bool succeeded =
-            placeholder_cache != nullptr && size >= 0 &&
-            (placeholder_cache->Exists(placeholder_id) ||
-             placeholder_cache->Create(placeholder_id, static_cast<u64>(size)));
-
-        if (succeeded) {
-            LOG_DEBUG(Service_NCM, "called, storage_id={}, size={}", static_cast<u32>(storage),
-                      size);
-        } else {
-            LOG_WARNING(Service_NCM, "failed, storage_id={}, size={}", static_cast<u32>(storage),
-                        size);
-        }
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(succeeded ? ResultSuccess : ResultUnknown);
-    }
-
-    void DeletePlaceHolder(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto placeholder_id = rp.PopRaw<FileSys::NcaID>();
-
-        auto* const placeholder_cache =
-            system.GetFileSystemController().GetPlaceholderCacheForStorage(storage);
-        const bool succeeded =
-            placeholder_cache != nullptr &&
-            (!placeholder_cache->Exists(placeholder_id) || placeholder_cache->Delete(placeholder_id));
-
-        if (succeeded) {
-            LOG_DEBUG(Service_NCM, "called, storage_id={}", static_cast<u32>(storage));
-        } else {
-            LOG_WARNING(Service_NCM, "failed, storage_id={}", static_cast<u32>(storage));
-        }
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(succeeded ? ResultSuccess : ResultUnknown);
-    }
-
-    void WritePlaceHolder(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto placeholder_id = rp.PopRaw<FileSys::NcaID>();
-        const auto offset = rp.Pop<u64>();
-        const auto data = ctx.ReadBuffer();
-
-        auto* const placeholder_cache =
-            system.GetFileSystemController().GetPlaceholderCacheForStorage(storage);
-        const std::vector<u8> write_data{data.begin(), data.end()};
-        const bool succeeded =
-            placeholder_cache != nullptr &&
-            placeholder_cache->Write(placeholder_id, offset, write_data);
-
-        if (succeeded) {
-            LOG_DEBUG(Service_NCM, "called, storage_id={}, offset={}, size={}",
-                      static_cast<u32>(storage), offset, data.size());
-        } else {
-            LOG_WARNING(Service_NCM, "failed, storage_id={}, offset={}, size={}",
-                        static_cast<u32>(storage), offset, data.size());
-        }
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(succeeded ? ResultSuccess : ResultUnknown);
-    }
-
-    void Register(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        FileSys::NcaID content_id{};
-        FileSys::NcaID placeholder_id{};
-        if constexpr (HLE::ApiVersion::HOS_VERSION_MAJOR >= 16) {
-            placeholder_id = rp.PopRaw<FileSys::NcaID>();
-            content_id = rp.PopRaw<FileSys::NcaID>();
-        } else {
-            content_id = rp.PopRaw<FileSys::NcaID>();
-            placeholder_id = rp.PopRaw<FileSys::NcaID>();
-        }
-
-        auto& fsc = system.GetFileSystemController();
-        auto* const placeholder_cache = fsc.GetPlaceholderCacheForStorage(storage);
-        auto* const registered_cache = fsc.GetRegisteredCacheForStorage(storage);
-        const bool succeeded =
-            placeholder_cache != nullptr && registered_cache != nullptr &&
-            placeholder_cache->Register(registered_cache, placeholder_id, content_id);
-
-        if (succeeded) {
-            LOG_DEBUG(Service_NCM, "called, storage_id={}", static_cast<u32>(storage));
-        } else {
-            LOG_WARNING(Service_NCM, "failed, storage_id={}", static_cast<u32>(storage));
-        }
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(succeeded ? ResultSuccess : ResultUnknown);
-    }
-
-    void Delete(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto content_id = rp.PopRaw<FileSys::NcaID>();
-
-        auto* const registered_cache =
-            system.GetFileSystemController().GetRegisteredCacheForStorage(storage);
-        const bool succeeded = registered_cache != nullptr && registered_cache->Delete(content_id);
-        if (succeeded) {
-            registered_cache->Refresh();
-        }
-
-        if (succeeded) {
-            LOG_DEBUG(Service_NCM, "called, storage_id={}", static_cast<u32>(storage));
-        } else {
-            LOG_WARNING(Service_NCM, "failed, storage_id={}", static_cast<u32>(storage));
-        }
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(succeeded ? ResultSuccess : ResultUnknown);
-    }
-
-    FileSys::StorageId storage;
-};
-
-class IContentMetaDatabase final : public ServiceFramework<IContentMetaDatabase> {
-public:
-    explicit IContentMetaDatabase(Core::System& system_, FileSys::StorageId id)
-        : ServiceFramework{system_, "IContentMetaDatabase"}, storage{id} {
-        // clang-format off
-        static const FunctionInfo functions[] = {
-            {0, &IContentMetaDatabase::Set, "Set"},
-            {2, &IContentMetaDatabase::Remove, "Remove"},
-            {8, &IContentMetaDatabase::Has, "Has"},
-            {15, &IContentMetaDatabase::Commit, "Commit"},
-        };
-        // clang-format on
-
-        RegisterHandlers(functions);
-    }
-
-private:
-    struct ContentMetaKey {
-        u64 id;
-        u32 version;
-        FileSys::TitleType type;
-        u8 install_type;
-        std::array<u8, 2> padding;
-    };
-    static_assert(sizeof(ContentMetaKey) == 0x10);
-
-    void Set(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto key = rp.PopRaw<ContentMetaKey>();
-
-        const auto entry_matches = [&key](const ContentMetaKey& entry) {
-            return entry.id == key.id && entry.version == key.version && entry.type == key.type &&
-                   entry.install_type == key.install_type;
-        };
-        if (std::find_if(entries.begin(), entries.end(), entry_matches) == entries.end()) {
-            entries.push_back(key);
-        }
-
-        LOG_DEBUG(Service_NCM,
-                  "called, storage_id={}, title_id={:016X}, version={}, type={}, size={}",
-                  static_cast<u32>(storage), key.id, key.version, static_cast<u8>(key.type),
-                  ctx.GetReadBufferSize());
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
-    }
-
-    void Remove(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto key = rp.PopRaw<ContentMetaKey>();
-
-        std::erase_if(entries, [&key](const ContentMetaKey& entry) {
-            return entry.id == key.id && entry.version == key.version && entry.type == key.type &&
-                   entry.install_type == key.install_type;
-        });
-
-        LOG_DEBUG(Service_NCM, "called, storage_id={}, title_id={:016X}, version={}, type={}",
-                  static_cast<u32>(storage), key.id, key.version, static_cast<u8>(key.type));
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
-    }
-
-    void Has(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto key = rp.PopRaw<ContentMetaKey>();
-
-        const bool has_pending =
-            std::find_if(entries.begin(), entries.end(), [&key](const ContentMetaKey& entry) {
-                return entry.id == key.id && entry.version == key.version &&
-                       entry.type == key.type && entry.install_type == key.install_type;
-            }) != entries.end();
-
-        auto* const registered_cache =
-            system.GetFileSystemController().GetRegisteredCacheForStorage(storage);
-        const bool has_registered =
-            registered_cache != nullptr &&
-            registered_cache->HasEntry(key.id, FileSys::ContentRecordType::Meta);
-
-        LOG_DEBUG(Service_NCM, "called, storage_id={}, title_id={:016X}, version={}, type={}, has={}",
-                  static_cast<u32>(storage), key.id, key.version, static_cast<u8>(key.type),
-                  has_pending || has_registered);
-
-        IPC::ResponseBuilder rb{ctx, 3};
-        rb.Push(ResultSuccess);
-        rb.Push(has_pending || has_registered);
-    }
-
-    void Commit(HLERequestContext& ctx) {
-        auto* const registered_cache =
-            system.GetFileSystemController().GetRegisteredCacheForStorage(storage);
-        if (registered_cache != nullptr) {
-            registered_cache->Refresh();
-        }
-
-        LOG_DEBUG(Service_NCM, "called, storage_id={}", static_cast<u32>(storage));
-
-        IPC::ResponseBuilder rb{ctx, 2};
-        rb.Push(ResultSuccess);
-    }
-
-    FileSys::StorageId storage;
-    std::vector<ContentMetaKey> entries;
-};
-
 class LR final : public ServiceFramework<LR> {
 public:
     explicit LR(Core::System& system_) : ServiceFramework{system_, "lr"} {
@@ -385,8 +113,8 @@ public:
             {1, nullptr, "CreateContentMetaDatabase"},
             {2, nullptr, "VerifyContentStorage"},
             {3, nullptr, "VerifyContentMetaDatabase"},
-            {4, &NCM::OpenContentStorage, "OpenContentStorage"},
-            {5, &NCM::OpenContentMetaDatabase, "OpenContentMetaDatabase"},
+            {4, nullptr, "OpenContentStorage"},
+            {5, nullptr, "OpenContentMetaDatabase"},
             {6, nullptr, "CloseContentStorageForcibly"},
             {7, nullptr, "CloseContentMetaDatabaseForcibly"},
             {8, nullptr, "CleanupContentMetaDatabase"},
@@ -402,29 +130,6 @@ public:
 
         RegisterHandlers(functions);
     }
-
-private:
-    void OpenContentStorage(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto storage_id = rp.PopEnum<FileSys::StorageId>();
-
-        LOG_DEBUG(Service_NCM, "called, storage_id={}", static_cast<u32>(storage_id));
-
-        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-        rb.Push(ResultSuccess);
-        rb.PushIpcInterface<IContentStorage>(ctx, system, storage_id);
-    }
-
-    void OpenContentMetaDatabase(HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-        const auto storage_id = rp.PopEnum<FileSys::StorageId>();
-
-        LOG_DEBUG(Service_NCM, "called, storage_id={}", static_cast<u32>(storage_id));
-
-        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-        rb.Push(ResultSuccess);
-        rb.PushIpcInterface<IContentMetaDatabase>(ctx, system, storage_id);
-    }
 };
 
 void LoopProcess(Core::System& system) {

src/core/hle/service/ns/application_manager_interface.cpp

@@ -9,7 +9,6 @@
 #include "core/file_sys/registered_cache.h"
 #include "core/hle/service/cmif_serialization.h"
 #include "core/hle/service/filesystem/filesystem.h"
-#include "core/hle/service/ipc_helpers.h"
 #include "core/hle/service/ns/application_manager_interface.h"
 
 #include "core/file_sys/content_archive.h"
@@ -20,7 +19,6 @@
 #include "core/launch_timestamp_cache.h"
 
 #include <algorithm>
-#include <cstring>
 #include <vector>
 
 namespace Service::NS {
@@ -38,14 +36,14 @@ IApplicationManagerInterface::IApplicationManagerInterface(Core::System& system_
         {1, nullptr, "GenerateApplicationRecordCount"},
         {2, D<&IApplicationManagerInterface::GetApplicationRecordUpdateSystemEvent>, "GetApplicationRecordUpdateSystemEvent"},
         {3, nullptr, "GetApplicationViewDeprecated"},
-        {4, D<&IApplicationManagerInterface::DeleteApplicationEntity>, "DeleteApplicationEntity"},
-        {5, D<&IApplicationManagerInterface::DeleteApplicationCompletely>, "DeleteApplicationCompletely"},
+        {4, nullptr, "DeleteApplicationEntity"},
+        {5, nullptr, "DeleteApplicationCompletely"},
         {6, nullptr, "IsAnyApplicationEntityRedundant"},
         {7, nullptr, "DeleteRedundantApplicationEntity"},
         {8, nullptr, "IsApplicationEntityMovable"},
         {9, nullptr, "MoveApplicationEntity"},
         {11, nullptr, "CalculateApplicationOccupiedSize"},
-        {16, &IApplicationManagerInterface::PushApplicationRecord, "PushApplicationRecord"},
+        {16, nullptr, "PushApplicationRecord"},
         {17, nullptr, "ListApplicationRecordContentMeta"},
         {19, nullptr, "LaunchApplicationOld"},
         {21, nullptr, "GetApplicationContentPath"},
@@ -645,27 +643,6 @@ Result IApplicationManagerInterface::IsAnyApplicationEntityInstalled(
     R_SUCCEED();
 }
 
-Result IApplicationManagerInterface::DeleteApplicationEntity(u64 application_id) {
-    LOG_DEBUG(Service_NS, "called, application_id={:016X}", application_id);
-
-    auto& fsc = system.GetFileSystemController();
-    if (auto* const user_cache = fsc.GetUserNANDContents(); user_cache != nullptr) {
-        user_cache->RemoveExistingEntry(application_id);
-        user_cache->Refresh();
-    }
-    if (auto* const sdmc_cache = fsc.GetSDMCContents(); sdmc_cache != nullptr) {
-        sdmc_cache->RemoveExistingEntry(application_id);
-        sdmc_cache->Refresh();
-    }
-
-    record_update_system_event.Signal(system.Kernel());
-    R_SUCCEED();
-}
-
-Result IApplicationManagerInterface::DeleteApplicationCompletely(u64 application_id) {
-    R_RETURN(DeleteApplicationEntity(application_id));
-}
-
 Result IApplicationManagerInterface::GetApplicationViewDeprecated(
     OutArray<ApplicationViewV19, BufferAttr_HipcMapAlias> out_application_views,
     InArray<u64, BufferAttr_HipcMapAlias> application_ids) {
@@ -866,29 +843,6 @@ Result IApplicationManagerInterface::Unknown4053() {
     R_SUCCEED();
 }
 
-void IApplicationManagerInterface::PushApplicationRecord(HLERequestContext& ctx) {
-    const auto record = ctx.ReadBuffer();
-    u64 application_id{};
-    if (record.size() >= sizeof(application_id)) {
-        std::memcpy(&application_id, record.data(), sizeof(application_id));
-    }
-
-    LOG_DEBUG(Service_NS, "called, application_id={:016X}, size={}", application_id, record.size());
-
-    auto& fsc = system.GetFileSystemController();
-    if (auto* const user_cache = fsc.GetUserNANDContents(); user_cache != nullptr) {
-        user_cache->Refresh();
-    }
-    if (auto* const sdmc_cache = fsc.GetSDMCContents(); sdmc_cache != nullptr) {
-        sdmc_cache->Refresh();
-    }
-
-    record_update_system_event.Signal(system.Kernel());
-
-    IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(ResultSuccess);
-}
-
 void IApplicationManagerInterface::ListApplicationTitle(HLERequestContext& ctx) {
     LOG_DEBUG(Service_NS, "called");
     IReadOnlyApplicationControlDataInterface(system).ListApplicationTitle(ctx);

src/core/hle/service/ns/application_manager_interface.h

@@ -56,8 +56,6 @@ public:
     Result ResumeAll();
     Result IsQualificationTransitionSupportedByProcessId(Out<bool> out_is_supported,
                                                          u64 process_id);
-    Result DeleteApplicationEntity(u64 application_id);
-    Result DeleteApplicationCompletely(u64 application_id);
     Result GetStorageSize(Out<s64> out_total_space_size, Out<s64> out_free_space_size,
                           FileSys::StorageId storage_id);
     Result TouchApplication(u64 application_id);
@@ -76,7 +74,6 @@ public:
     Result RequestDownloadApplicationControlDataInBackground(u64 control_source,
                                                              u64 application_id);
 
-    void PushApplicationRecord(HLERequestContext& ctx);
     void ListApplicationTitle(HLERequestContext& ctx);
 
 private:

src/core/hle/service/spl/spl.cpp

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -31,9 +28,9 @@ SPL_MIG::SPL_MIG(Core::System& system_, std::shared_ptr<Module> module_)
     static const FunctionInfo functions[] = {
         {0, &SPL::GetConfig, "GetConfig"},
         {1, &SPL::ModularExponentiate, "ModularExponentiate"},
-        {2, &SPL::GenerateAesKek, "GenerateAesKek"},
+        {2, nullptr, "GenerateAesKek"},
         {3, nullptr, "LoadAesKey"},
-        {4, &SPL::GenerateAesKey, "GenerateAesKey"},
+        {4, nullptr, "GenerateAesKey"},
         {5, &SPL::SetConfig, "SetConfig"},
         {7, &SPL::GenerateRandomBytes, "GenerateRandomBytes"},
         {11, &SPL::IsDevelopment, "IsDevelopment"},

src/core/hle/service/spl/spl_module.cpp

@@ -59,36 +59,6 @@ void Module::Interface::ModularExponentiate(HLERequestContext& ctx) {
     rb.Push(ResultSecureMonitorNotImplemented);
 }
 
-void Module::Interface::GenerateAesKek(HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    [[maybe_unused]] const auto key_source = rp.PopRaw<KeySource>();
-    const auto generation = rp.Pop<u32>();
-    const auto option = rp.Pop<u32>();
-
-    LOG_WARNING(Service_SPL, "(STUBBED) called, generation={:#x}, option={:#x}", generation,
-                option);
-
-    AccessKey access_key{};
-
-    IPC::ResponseBuilder rb{ctx, 6};
-    rb.Push(ResultSuccess);
-    rb.PushRaw(access_key);
-}
-
-void Module::Interface::GenerateAesKey(HLERequestContext& ctx) {
-    IPC::RequestParser rp{ctx};
-    [[maybe_unused]] const auto access_key = rp.PopRaw<AccessKey>();
-    [[maybe_unused]] const auto key_source = rp.PopRaw<KeySource>();
-
-    LOG_WARNING(Service_SPL, "(STUBBED) called");
-
-    AesKey aes_key{};
-
-    IPC::ResponseBuilder rb{ctx, 6};
-    rb.Push(ResultSuccess);
-    rb.PushRaw(aes_key);
-}
-
 void Module::Interface::SetConfig(HLERequestContext& ctx) {
     UNIMPLEMENTED_MSG("SetConfig is not implemented!");
 

src/core/hle/service/spl/spl_module.h

@@ -1,6 +1,3 @@
-// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
-// SPDX-License-Identifier: GPL-3.0-or-later
-
 // SPDX-FileCopyrightText: Copyright 2018 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -28,8 +25,6 @@ public:
         // General
         void GetConfig(HLERequestContext& ctx);
         void ModularExponentiate(HLERequestContext& ctx);
-        void GenerateAesKek(HLERequestContext& ctx);
-        void GenerateAesKey(HLERequestContext& ctx);
         void SetConfig(HLERequestContext& ctx);
         void GenerateRandomBytes(HLERequestContext& ctx);
         void IsDevelopment(HLERequestContext& ctx);

src/video_core/buffer_cache/buffer_cache.h

@@ -121,6 +121,15 @@ void BufferCache<P>::WriteMemory(DAddr device_addr, u64 size) {
     memory_tracker.MarkRegionAsCpuModified(device_addr, size);
 }
 
+template <class P>
+void BufferCache<P>::CpuWriteInvalidate(DAddr device_addr, u64 size) {
+    if (!memory_tracker.CpuMarkIfNotGpuModified(device_addr, size)) {
+        return;
+    }
+    std::scoped_lock lock{mutex};
+    WriteMemory(device_addr, size);
+}
+
 template <class P>
 void BufferCache<P>::CachedWriteMemory(DAddr device_addr, u64 size) {
     const bool is_dirty = IsRegionRegistered(device_addr, size);
@@ -175,9 +184,71 @@ std::optional<VideoCore::RasterizerDownloadArea> BufferCache<P>::GetFlushArea(DA
 
 template <class P>
 void BufferCache<P>::DownloadMemory(DAddr device_addr, u64 size) {
-    ForEachBufferInRange(device_addr, size, [&](BufferId, Buffer& buffer) {
-        DownloadBufferMemory(buffer, device_addr, size);
+    if constexpr (!USE_MEMORY_MAPS) {
+        std::scoped_lock lock{mutex};
+        ForEachBufferInRange(device_addr, size, [&](BufferId, Buffer& buffer) {
+            DownloadBufferMemory(buffer, device_addr, size);
+        });
+        return;
+    }
+
+    boost::container::small_vector<std::pair<BufferCopy, BufferId>, 8> downloads;
+    u64 total_size_bytes = 0;
+    u64 largest_copy = 0;
+
+    std::unique_lock lock{mutex};
+    ForEachBufferInRange(device_addr, size, [&](BufferId buffer_id, Buffer& buffer) {
+        memory_tracker.ForEachDownloadRangeAndClear(
+            device_addr, size, [&](u64 device_addr_out, u64 range_size) {
+                const DAddr buffer_addr = buffer.CpuAddr();
+                const auto add_download = [&](DAddr start, DAddr end) {
+                    const u64 new_offset = start - buffer_addr;
+                    const u64 new_size = end - start;
+                    downloads.push_back({
+                        BufferCopy{
+                            .src_offset = new_offset,
+                            .dst_offset = total_size_bytes,
+                            .size = new_size,
+                        },
+                        buffer_id,
+                    });
+                    constexpr u64 align = 64ULL;
+                    constexpr u64 mask = ~(align - 1ULL);
+                    total_size_bytes += (new_size + align - 1) & mask;
+                    largest_copy = (std::max)(largest_copy, new_size);
+                };
+                gpu_modified_ranges.ForEachInRange(device_addr_out, range_size, add_download);
+                ClearDownload(device_addr_out, range_size);
+                gpu_modified_ranges.Subtract(device_addr_out, range_size);
+            });
     });
+    if (total_size_bytes == 0) {
+        return;
+    }
+
+    auto download_staging = runtime.DownloadStagingBuffer(total_size_bytes);
+    boost::container::small_vector<BufferCopy, 8> writebacks;
+    runtime.PreCopyBarrier();
+    for (auto& [copy, buffer_id] : downloads) {
+        copy.dst_offset += download_staging.offset;
+        Buffer& buffer = slot_buffers[buffer_id];
+        buffer.MarkUsage(copy.src_offset, copy.size);
+        const std::array copies{copy};
+        runtime.CopyBuffer(download_staging.buffer, buffer, copies, false);
+        BufferCopy writeback{copy};
+        writeback.src_offset = static_cast<u64>(buffer.CpuAddr()) + copy.src_offset;
+        writebacks.push_back(writeback);
+    }
+    runtime.PostCopyBarrier();
+    lock.unlock();
+
+    runtime.Finish();
+    const u8* const base = download_staging.mapped_span.data();
+    for (const BufferCopy& writeback : writebacks) {
+        const u64 staging_offset = writeback.dst_offset - download_staging.offset;
+        device_memory.WriteBlockUnsafe(static_cast<DAddr>(writeback.src_offset),
+                                       base + staging_offset, writeback.size);
+    }
 }
 
 template <class P>
@@ -214,7 +285,7 @@ bool BufferCache<P>::DMACopy(GPUVAddr src_address, GPUVAddr dest_address, u64 am
     auto& src_buffer = slot_buffers[buffer_a];
     auto& dest_buffer = slot_buffers[buffer_b];
     SynchronizeBuffer(src_buffer, *cpu_src_address, static_cast<u32>(amount));
-    SynchronizeBuffer(dest_buffer, *cpu_dest_address, static_cast<u32>(amount));
+    memory_tracker.UnmarkRegionAsCpuModified(*cpu_dest_address, static_cast<u32>(amount));
     std::array copies{BufferCopy{
         .src_offset = src_buffer.Offset(*cpu_src_address),
         .dst_offset = dest_buffer.Offset(*cpu_dest_address),
@@ -673,32 +744,44 @@ void BufferCache<P>::PopAsyncFlushes() {
 
 template <class P>
 void BufferCache<P>::PopAsyncBuffers() {
-    if (async_buffers.empty()) {
-        return;
-    }
-    if (!async_buffers.front().has_value()) {
+    struct Writeback {
+        DAddr addr;
+        const u8* src;
+        u64 size;
+    };
+    boost::container::small_vector<Writeback, 8> writebacks;
+    {
+        std::scoped_lock lock{mutex};
+        if (async_buffers.empty()) {
+            return;
+        }
+        if (!async_buffers.front().has_value()) {
+            async_buffers.pop_front();
+            return;
+        }
+        auto& downloads = pending_downloads.front();
+        auto& async_buffer = async_buffers.front();
+        const u8* base = async_buffer->mapped_span.data();
+        const size_t base_offset = async_buffer->offset;
+        for (const auto& copy : downloads) {
+            const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
+            const u64 dst_offset = copy.dst_offset - base_offset;
+            const u8* read_mapped_memory = base + dst_offset;
+            async_downloads.ForEachInRange(device_addr, copy.size, [&](DAddr start, DAddr end, s32) {
+                writebacks.push_back(
+                    {start, &read_mapped_memory[start - device_addr], end - start});
+            });
+            async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
+                gpu_modified_ranges.Subtract(start, end - start);
+            });
+        }
+        async_buffers_death_ring.emplace_back(*async_buffer);
         async_buffers.pop_front();
-        return;
+        pending_downloads.pop_front();
     }
-    auto& downloads = pending_downloads.front();
-    auto& async_buffer = async_buffers.front();
-    u8* base = async_buffer->mapped_span.data();
-    const size_t base_offset = async_buffer->offset;
-    for (const auto& copy : downloads) {
-        const DAddr device_addr = static_cast<DAddr>(copy.src_offset);
-        const u64 dst_offset = copy.dst_offset - base_offset;
-        const u8* read_mapped_memory = base + dst_offset;
-        async_downloads.ForEachInRange(device_addr, copy.size, [&](DAddr start, DAddr end, s32) {
-            device_memory.WriteBlockUnsafe(start, &read_mapped_memory[start - device_addr],
-                                           end - start);
-        });
-        async_downloads.Subtract(device_addr, copy.size, [&](DAddr start, DAddr end) {
-            gpu_modified_ranges.Subtract(start, end - start);
-        });
+    for (const auto& wb : writebacks) {
+        device_memory.WriteBlockUnsafe(wb.addr, wb.src, wb.size);
     }
-    async_buffers_death_ring.emplace_back(*async_buffer);
-    async_buffers.pop_front();
-    pending_downloads.pop_front();
 }
 
 template <class P>
@@ -1638,6 +1721,9 @@ void BufferCache<P>::TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept {
 
 template <class P>
 bool BufferCache<P>::SynchronizeBuffer(Buffer& buffer, DAddr device_addr, u32 size) {
+    if (!memory_tracker.HasCpuModifiedCheap(device_addr, size)) {
+        return true;
+    }
     upload_copies.clear();
     u64 total_size_bytes = 0;
     u64 largest_copy = 0;

src/video_core/buffer_cache/buffer_cache_base.h

@@ -217,6 +217,8 @@ public:
 
     void WriteMemory(DAddr device_addr, u64 size);
 
+    void CpuWriteInvalidate(DAddr device_addr, u64 size);
+
     void CachedWriteMemory(DAddr device_addr, u64 size);
 
     bool OnCPUWrite(DAddr device_addr, u64 size);

src/video_core/buffer_cache/memory_tracker_base.h

@@ -7,9 +7,11 @@
 #pragma once
 
 #include <algorithm>
+#include <atomic>
 #include <bit>
 #include <deque>
 #include <limits>
+#include <mutex>
 #include <type_traits>
 #include <ankerl/unordered_dense.h>
 #include <utility>
@@ -49,6 +51,24 @@ public:
         });
     }
 
+    [[nodiscard]] bool HasCpuModifiedCheap(VAddr query_cpu_addr, u64 query_size) noexcept {
+        std::size_t remaining_size{query_size};
+        std::size_t page_index{query_cpu_addr >> HIGHER_PAGE_BITS};
+        u64 page_offset{query_cpu_addr & HIGHER_PAGE_MASK};
+        while (remaining_size > 0) {
+            const std::size_t copy_amount{
+                std::min<std::size_t>(HIGHER_PAGE_SIZE - page_offset, remaining_size)};
+            const Manager* manager = top_tier[page_index].load(std::memory_order_acquire);
+            if (manager == nullptr || manager->CpuModifiedPageCount() != 0) {
+                return true;
+            }
+            page_index++;
+            page_offset = 0;
+            remaining_size -= copy_amount;
+        }
+        return false;
+    }
+
     /// Returns true if a region has been modified from the CPU
     [[nodiscard]] bool IsRegionCpuModified(VAddr query_cpu_addr, u64 query_size) noexcept {
         return IteratePages<true>(query_cpu_addr, query_size, [](Manager* manager, u64 offset, size_t size) {
@@ -130,12 +150,28 @@ public:
     }
 
     void FlushCachedWrites() noexcept {
+        std::scoped_lock lk{tracker_mutex};
         for (auto id : cached_pages) {
-            top_tier[id]->FlushCachedWrites();
+            top_tier[id].load(std::memory_order_relaxed)->FlushCachedWrites();
         }
         cached_pages.clear();
     }
 
+    [[nodiscard]] bool CpuMarkIfNotGpuModified(VAddr addr, u64 size) {
+        std::scoped_lock lk{tracker_mutex};
+        const bool gpu = IteratePagesNoLock<false>(
+            addr, size, [](Manager* manager, u64 offset, size_t sz) {
+                return manager->IsRegionModified(Type::GPU, offset, sz);
+            });
+        if (gpu) {
+            return true;
+        }
+        IteratePagesNoLock<true>(addr, size, [](Manager* manager, u64 offset, size_t sz) {
+            manager->ChangeRegionState(Type::CPU, true, manager->cpu_addr + offset, sz);
+        });
+        return false;
+    }
+
     /// Call 'func' for each CPU modified range and unmark those pages as CPU modified
     template <typename Func>
     void ForEachUploadRange(VAddr query_cpu_range, u64 query_size, Func&& func) {
@@ -162,6 +198,12 @@ public:
 private:
     template <bool create_region_on_fail, typename Func>
     bool IteratePages(VAddr cpu_address, size_t size, Func&& func) {
+        std::scoped_lock lk{tracker_mutex};
+        return IteratePagesNoLock<create_region_on_fail>(cpu_address, size, std::forward<Func>(func));
+    }
+
+    template <bool create_region_on_fail, typename Func>
+    bool IteratePagesNoLock(VAddr cpu_address, size_t size, Func&& func) {
         using FuncReturn = typename std::invoke_result<Func, Manager*, u64, size_t>::type;
         static constexpr bool BOOL_BREAK = std::is_same_v<FuncReturn, bool>;
         std::size_t remaining_size{size};
@@ -170,7 +212,7 @@ private:
         while (remaining_size > 0) {
             const std::size_t copy_amount{
                 std::min<std::size_t>(HIGHER_PAGE_SIZE - page_offset, remaining_size)};
-            auto* manager{top_tier[page_index]};
+            auto* manager{top_tier[page_index].load(std::memory_order_relaxed)};
             if (manager) {
                 if constexpr (BOOL_BREAK) {
                     if (func(manager, page_offset, copy_amount)) {
@@ -181,7 +223,7 @@ private:
                 }
             } else if constexpr (create_region_on_fail) {
                 CreateRegion(page_index);
-                manager = top_tier[page_index];
+                manager = top_tier[page_index].load(std::memory_order_relaxed);
                 if constexpr (BOOL_BREAK) {
                     if (func(manager, page_offset, copy_amount)) {
                         return true;
@@ -199,6 +241,7 @@ private:
 
     template <bool create_region_on_fail, typename Func>
     std::pair<u64, u64> IteratePairs(VAddr cpu_address, size_t size, Func&& func) {
+        std::scoped_lock lk{tracker_mutex};
         std::size_t remaining_size{size};
         std::size_t page_index{cpu_address >> HIGHER_PAGE_BITS};
         u64 page_offset{cpu_address & HIGHER_PAGE_MASK};
@@ -207,7 +250,7 @@ private:
         while (remaining_size > 0) {
             const std::size_t copy_amount{
                 std::min<std::size_t>(HIGHER_PAGE_SIZE - page_offset, remaining_size)};
-            auto* manager{top_tier[page_index]};
+            auto* manager{top_tier[page_index].load(std::memory_order_relaxed)};
             const auto execute = [&] {
                 auto [new_begin, new_end] = func(manager, page_offset, copy_amount);
                 if (new_begin != 0 || new_end != 0) {
@@ -220,7 +263,7 @@ private:
                 execute();
             } else if constexpr (create_region_on_fail) {
                 CreateRegion(page_index);
-                manager = top_tier[page_index];
+                manager = top_tier[page_index].load(std::memory_order_relaxed);
                 execute();
             }
             page_index++;
@@ -236,7 +279,7 @@ private:
 
     void CreateRegion(std::size_t page_index) {
         const VAddr base_cpu_addr = page_index << HIGHER_PAGE_BITS;
-        top_tier[page_index] = GetNewManager(base_cpu_addr);
+        top_tier[page_index].store(GetNewManager(base_cpu_addr), std::memory_order_release);
     }
 
     Manager* GetNewManager(VAddr base_cpu_address) {
@@ -254,11 +297,12 @@ private:
         return new_manager;
     }
 
-    std::array<Manager*, NUM_HIGH_PAGES> top_tier{};
+    std::array<std::atomic<Manager*>, NUM_HIGH_PAGES> top_tier{};
     std::deque<std::array<Manager, MANAGER_POOL_SIZE>> manager_pool;
     std::deque<Manager*> free_managers;
     ankerl::unordered_dense::set<u32> cached_pages;
     DeviceTracker* device_tracker = nullptr;
+    std::mutex tracker_mutex;
 };
 
 } // namespace VideoCommon

src/video_core/buffer_cache/word_manager.h

@@ -7,6 +7,7 @@
 #pragma once
 
 #include <algorithm>
+#include <atomic>
 #include <bit>
 #include <limits>
 #include <span>
@@ -48,6 +49,11 @@ struct WordManager {
         u64 const last_word = (~u64{0} << shift) >> shift;
         heap[num_words * size_t(Type::CPU) + num_words - 1] = last_word;
         heap[num_words * size_t(Type::Untracked) + num_words - 1] = last_word;
+        u32 cpu_pages = 0;
+        for (size_t i = 0; i < num_words; ++i) {
+            cpu_pages += static_cast<u32>(std::popcount(heap[num_words * size_t(Type::CPU) + i]));
+        }
+        cpu_modified_pages.store(cpu_pages, std::memory_order_relaxed);
     }
     explicit WordManager() = default;
 
@@ -120,10 +126,15 @@ struct WordManager {
         [[maybe_unused]] std::span<u64> untracked_words = Span(Type::Untracked);
         [[maybe_unused]] std::span<u64> cached_words = Span(Type::CachedCPU);
         std::vector<std::pair<VAddr, u64>> ranges;
+        s64 cpu_delta = 0;
         IterateWords(dirty_addr - cpu_addr, size, [&](size_t index, u64 mask) {
             if (type == Type::CPU || type == Type::CachedCPU) {
                 CollectChangedRanges(!enable, index, untracked_words[index], mask, ranges);
             }
+            if (type == Type::CPU) {
+                const u64 old = state_words[index];
+                cpu_delta += enable ? std::popcount(~old & mask) : -std::popcount(old & mask);
+            }
             if (enable) {
                 state_words[index] |= mask;
                 if (type == Type::CPU || type == Type::CachedCPU)
@@ -138,6 +149,9 @@ struct WordManager {
                     untracked_words[index] &= ~mask;
             }
         });
+        if (cpu_delta != 0) {
+            cpu_modified_pages.fetch_add(static_cast<u32>(cpu_delta), std::memory_order_release);
+        }
         if (!ranges.empty()) {
             ApplyCollectedRanges(ranges, (!enable) ? 1 : -1);
         }
@@ -165,6 +179,7 @@ struct WordManager {
                  (pending_pointer - pending_offset) * BYTES_PER_PAGE);
         };
         std::vector<std::pair<VAddr, u64>> ranges;
+        s64 cpu_delta = 0;
         IterateWords(offset, size, [&](size_t index, u64 mask) {
             if (type == Type::GPU)
                 mask &= ~untracked_words[index];
@@ -173,6 +188,8 @@ struct WordManager {
                 if (type == Type::CPU || type == Type::CachedCPU) {
                     CollectChangedRanges(true, index, untracked_words[index], mask, ranges);
                 }
+                if (type == Type::CPU)
+                    cpu_delta -= std::popcount(word);
                 state_words[index] &= ~mask;
                 if (type == Type::CPU || type == Type::CachedCPU)
                     untracked_words[index] &= ~mask;
@@ -194,6 +211,9 @@ struct WordManager {
                 }
             });
         });
+        if (cpu_delta != 0) {
+            cpu_modified_pages.fetch_add(static_cast<u32>(cpu_delta), std::memory_order_release);
+        }
         if (pending) {
             release();
         }
@@ -248,13 +268,18 @@ struct WordManager {
         auto const untracked_words = Span(Type::Untracked);
         auto const cpu_words = Span(Type::CPU);
         std::vector<std::pair<VAddr, u64>> ranges;
+        s64 cpu_delta = 0;
         for (u64 word_index = 0; word_index < num_words; ++word_index) {
             const u64 cached_bits = cached_words[word_index];
             CollectChangedRanges(false, word_index, untracked_words[word_index], cached_bits, ranges);
+            cpu_delta += std::popcount(~cpu_words[word_index] & cached_bits);
             untracked_words[word_index] |= cached_bits;
             cpu_words[word_index] |= cached_bits;
             cached_words[word_index] = 0;
         }
+        if (cpu_delta != 0) {
+            cpu_modified_pages.fetch_add(static_cast<u32>(cpu_delta), std::memory_order_release);
+        }
         if (!ranges.empty()) {
             ApplyCollectedRanges(ranges, -1);
         }
@@ -319,9 +344,14 @@ struct WordManager {
         return std::span<const u64>(heap.data() + num_words * size_t(type), num_words);
     }
 
+    [[nodiscard]] u32 CpuModifiedPageCount() const noexcept {
+        return cpu_modified_pages.load(std::memory_order_acquire);
+    }
+
     std::array<u64, size_t(Type::Max) * num_words> heap = {};
     DeviceTracker* tracker = nullptr;
     VAddr cpu_addr = 0;
+    std::atomic<u32> cpu_modified_pages{0};
 };
 
 } // namespace VideoCommon

src/video_core/control/channel_state.cpp

@@ -16,26 +16,6 @@
 #include "video_core/memory_manager.h"
 
 namespace Tegra::Control {
-namespace {
-
-// Match NVK/Nouveau's initial pushbuffer subchannel layout.
-constexpr u32 Nvk3DSubchannel = 0;
-constexpr u32 NvkComputeSubchannel = 1;
-constexpr u32 Nvk2DSubchannel = 3;
-constexpr u32 NvkCopySubchannel = 4;
-
-void BindNvkDefaultSubchannels(ChannelState::Payload& payload) {
-    auto& dma_pusher = payload.dma_pusher;
-    dma_pusher.BindSubchannel(&payload.maxwell_3d, Nvk3DSubchannel, Engines::EngineTypes::Maxwell3D);
-    dma_pusher.BindSubchannel(&payload.kepler_compute, NvkComputeSubchannel,
-                              Engines::EngineTypes::KeplerCompute);
-    // Subchannel 2 is M2MF there; Eden does not expose a 0x9039 engine yet.
-    dma_pusher.BindSubchannel(&payload.fermi_2d, Nvk2DSubchannel, Engines::EngineTypes::Fermi2D);
-    dma_pusher.BindSubchannel(&payload.maxwell_dma, NvkCopySubchannel,
-                              Engines::EngineTypes::MaxwellDMA);
-}
-
-} // Anonymous namespace
 
 ChannelState::Payload::Payload(Core::System& system, MemoryManager& memory_manager, ChannelState& channel_state)
     : maxwell_3d(memory_manager)
@@ -55,7 +35,6 @@ void ChannelState::Init(Core::System& system, u64 program_id_) {
     ASSERT(memory_manager);
     program_id = program_id_;
     payload.emplace(system, *memory_manager, *this);
-    BindNvkDefaultSubchannels(*payload);
     initialized = true;
 }
 

src/video_core/engines/puller.cpp

@@ -22,21 +22,10 @@
 
 namespace Tegra::Engines {
 
-namespace {
-
-constexpr u32 Gf100BindClassMask = 0xffff;
-constexpr u32 Gf100BindValidMask =  0x1f0000 | Gf100BindClassMask;
-
-} // Anonymous namespace
-
 void Puller::ProcessBindMethod(DmaPusher& dma_pusher, const MethodCall& method_call) {
-    LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {:#x}", method_call.subchannel, method_call.argument);
-    u32 engine = method_call.argument;
-    if ((engine & ~Gf100BindClassMask) != 0 && (engine & ~Gf100BindValidMask) == 0) {
-        engine &= Gf100BindClassMask;
-    }
-
-    const auto engine_id = static_cast<EngineID>(engine);
+    // Bind the current subchannel to the desired engine id.
+    LOG_DEBUG(HW_GPU, "Binding subchannel {} to engine {}", method_call.subchannel, method_call.argument);
+    const auto engine_id = static_cast<EngineID>(method_call.argument);
     bound_engines[method_call.subchannel] = engine_id;
     switch (engine_id) {
     case EngineID::FERMI_TWOD_A:

src/video_core/fence_manager.h

@@ -91,9 +91,6 @@ public:
             func();
         }
         fences.push(std::move(new_fence));
-        if (should_flush) {
-            rasterizer.FlushCommands();
-        }
         if constexpr (can_async_check) {
             guard.unlock();
             cv.notify_all();
@@ -238,10 +235,10 @@ private:
 
     void PopAsyncFlushes() {
         {
-            std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+            std::scoped_lock lock{texture_cache.mutex};
             texture_cache.PopAsyncFlushes();
-            buffer_cache.PopAsyncFlushes();
         }
+        buffer_cache.PopAsyncFlushes();
         query_cache.PopAsyncFlushes();
     }
 

src/video_core/host_shaders/CMakeLists.txt

@@ -15,6 +15,8 @@ set(GLSL_INCLUDES
 
 set(SHADER_FILES
     ${CMAKE_CURRENT_SOURCE_DIR}/astc_decoder.comp
+    ${CMAKE_CURRENT_SOURCE_DIR}/astc_decoder.frag
+    ${CMAKE_CURRENT_SOURCE_DIR}/astc_decoder.vert
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_color_float.frag
     ${CMAKE_CURRENT_SOURCE_DIR}/block_linear_unswizzle_2d.comp
     ${CMAKE_CURRENT_SOURCE_DIR}/blit_color_msaa.frag

src/video_core/host_shaders/astc_decoder.comp

@@ -964,35 +964,70 @@ void ComputeEndpoints(out uvec4 ep1, out uvec4 ep2, uint color_endpoint_mode, ui
 }
 
 uint UnquantizeTexelWeight(EncodingData val) {
-    uint encoding = Encoding(val), bitlen = NumBits(val), bitval = BitValue(val);
-    if (encoding == JUST_BITS) {
-        return (bitlen >= 1 && bitlen <= 5)
-            ? uint(floor(0.5f + float(bitval) * 64.0f / float((1 << bitlen) - 1)))
-            : FastReplicateTo6(bitval, bitlen);
-    } else if (encoding == TRIT || encoding == QUINT) {
-        uint B = 0, C = 0, D = 0;
-        uint b_mask = (0x3100 >> (bitlen * 4)) & 0xf;
-        uint b = (bitval >> 1) & b_mask;
+    const uint encoding = Encoding(val);
+    const uint bitlen = NumBits(val);
+    const uint bitval = BitValue(val);
+    const uint A = ReplicateBitTo7((bitval & 1));
+    uint B = 0, C = 0, D = 0;
+    uint result = 0;
+    const uint bitlen_0_results[5] = {0, 16, 32, 48, 64};
+    switch (encoding) {
+    case JUST_BITS:
+        return FastReplicateTo6(bitval, bitlen);
+    case TRIT: {
         D = QuintTritValue(val);
-        if (encoding == TRIT) {
-            switch (bitlen) {
-            case 0: return D * 32; //0,32,64
-            case 1: C = 50; break;
-            case 2: C = 23; B = (b << 6) | (b << 2) | b; break;
-            case 3: C = 11; B = (b << 5) | b; break;
-            }
-        } else if (encoding == QUINT) {
-            switch (bitlen) {
-            case 0: return D * 16; //0, 16, 32, 48, 64
-            case 1: C = 28; break;
-            case 2: C = 13; B = (b << 6) | (b << 1); break;
-            }
+        switch (bitlen) {
+        case 0:
+            return bitlen_0_results[D * 2];
+        case 1: {
+            C = 50;
+            break;
         }
-        uint A = ReplicateBitTo7(bitval & 1);
-        uint res = (A & 0x20) | (((D * C + B) ^ A) >> 2);
-        return res + (res > 32 ? 1 : 0);
+        case 2: {
+            C = 23;
+            const uint b = (bitval >> 1) & 1;
+            B = (b << 6) | (b << 2) | b;
+            break;
+        }
+        case 3: {
+            C = 11;
+            const uint cb = (bitval >> 1) & 3;
+            B = (cb << 5) | cb;
+            break;
+        }
+        default:
+            break;
+        }
+        break;
     }
-    return 0;
+    case QUINT: {
+        D = QuintTritValue(val);
+        switch (bitlen) {
+        case 0:
+            return bitlen_0_results[D];
+        case 1: {
+            C = 28;
+            break;
+        }
+        case 2: {
+            C = 13;
+            const uint b = (bitval >> 1) & 1;
+            B = (b << 6) | (b << 1);
+            break;
+        }
+        }
+        break;
+    }
+    }
+    if (encoding != JUST_BITS && bitlen > 0) {
+        result = D * C + B;
+        result ^= A;
+        result = (A & 0x20) | (result >> 2);
+    }
+    if (result > 32) {
+        result += 1;
+    }
+    return result;
 }
 
 void UnquantizeTexelWeights(uvec2 size, bool is_dual_plane) {
@@ -1394,11 +1429,10 @@ void DecompressBlock(ivec3 coord) {
 }
 
 uint SwizzleOffset(uvec2 pos) {
-    return ((pos.x & 32u) << 3u) |
-           ((pos.y & 6u)  << 5u) |
-           ((pos.x & 16u) << 1u) |
-           ((pos.y & 1u)  << 4u) |
-           (pos.x & 15u);
+    const uint x = pos.x;
+    const uint y = pos.y;
+    return ((x % 64) / 32) * 256 + ((y % 8) / 2) * 64 +
+            ((x % 32) / 16) * 32 + (y % 2) * 16 + (x % 16);
 }
 
 void main() {

src/video_core/host_shaders/astc_decoder.vert

@@ -0,0 +1,19 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#version 450
+
+#ifdef VULKAN
+#define VERTEX_ID gl_VertexIndex
+#else
+#define VERTEX_ID gl_VertexID
+out gl_PerVertex {
+    vec4 gl_Position;
+};
+#endif
+
+void main() {
+    float x = float((VERTEX_ID & 1) << 2);
+    float y = float((VERTEX_ID & 2) << 1);
+    gl_Position = vec4(x - 1.0, y - 1.0, 0.0, 1.0);
+}

src/video_core/memory_manager.cpp

@@ -58,8 +58,9 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, GPU
 
 MemoryManager::~MemoryManager() = default;
 
-MemoryManager::EntryType MemoryManager::GetEntry(size_t position, bool is_big_page) const {
-    if (is_big_page) {
+template <bool is_big_page>
+MemoryManager::EntryType MemoryManager::GetEntry(size_t position) const {
+    if constexpr (is_big_page) {
         position = position >> big_page_bits;
         const u64 entry_mask = big_entries[position / 32];
         const size_t sub_index = position % 32;
@@ -72,8 +73,9 @@ MemoryManager::EntryType MemoryManager::GetEntry(size_t position, bool is_big_pa
     }
 }
 
-void MemoryManager::SetEntry(size_t position, MemoryManager::EntryType entry, bool is_big_page) {
-    if (is_big_page) {
+template <bool is_big_page>
+void MemoryManager::SetEntry(size_t position, MemoryManager::EntryType entry) {
+    if constexpr (is_big_page) {
         position = position >> big_page_bits;
         const u64 entry_mask = big_entries[position / 32];
         const size_t sub_index = position % 32;
@@ -106,21 +108,23 @@ inline void MemoryManager::SetBigPageContinuous(size_t big_page_index, bool valu
         (~(1ULL << sub_index) & continuous_mask) | (value ? 1ULL << sub_index : 0);
 }
 
-GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size, PTEKind kind, MemoryManager::EntryType entry_type) {
+template <MemoryManager::EntryType entry_type>
+GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size,
+                                    PTEKind kind) {
     [[maybe_unused]] u64 remaining_size{size};
-    if (entry_type == EntryType::Mapped) {
+    if constexpr (entry_type == EntryType::Mapped) {
         page_table.ReserveRange(gpu_addr, size);
     }
     for (u64 offset{}; offset < size; offset += page_size) {
         const GPUVAddr current_gpu_addr = gpu_addr + offset;
-        [[maybe_unused]] const auto current_entry_type = GetEntry(current_gpu_addr, false);
-        SetEntry(current_gpu_addr, entry_type, false);
+        [[maybe_unused]] const auto current_entry_type = GetEntry<false>(current_gpu_addr);
+        SetEntry<false>(current_gpu_addr, entry_type);
         if (current_entry_type != entry_type) {
             rasterizer->ModifyGPUMemory(unique_identifier, current_gpu_addr, page_size);
         }
-        if (entry_type == EntryType::Mapped) {
+        if constexpr (entry_type == EntryType::Mapped) {
             const DAddr current_dev_addr = dev_addr + offset;
-            const auto index = PageEntryIndex(current_gpu_addr, false);
+            const auto index = PageEntryIndex<false>(current_gpu_addr);
             const u32 sub_value = static_cast<u32>(current_dev_addr >> cpu_page_bits);
             page_table[index] = sub_value;
         }
@@ -130,18 +134,20 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr de
     return gpu_addr;
 }
 
-GPUVAddr MemoryManager::BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size, PTEKind kind, MemoryManager::EntryType entry_type) {
+template <MemoryManager::EntryType entry_type>
+GPUVAddr MemoryManager::BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr,
+                                       size_t size, PTEKind kind) {
     [[maybe_unused]] u64 remaining_size{size};
     for (u64 offset{}; offset < size; offset += big_page_size) {
         const GPUVAddr current_gpu_addr = gpu_addr + offset;
-        [[maybe_unused]] const auto current_entry_type = GetEntry(current_gpu_addr, true);
-        SetEntry(current_gpu_addr, entry_type, true);
+        [[maybe_unused]] const auto current_entry_type = GetEntry<true>(current_gpu_addr);
+        SetEntry<true>(current_gpu_addr, entry_type);
         if (current_entry_type != entry_type) {
             rasterizer->ModifyGPUMemory(unique_identifier, current_gpu_addr, big_page_size);
         }
-        if (entry_type == EntryType::Mapped) {
+        if constexpr (entry_type == EntryType::Mapped) {
             const DAddr current_dev_addr = dev_addr + offset;
-            const auto index = PageEntryIndex(current_gpu_addr, true);
+            const auto index = PageEntryIndex<true>(current_gpu_addr);
             const u32 sub_value = static_cast<u32>(current_dev_addr >> cpu_page_bits);
             big_page_table_dev[index] = sub_value;
             const bool is_continuous = ([&] {
@@ -175,16 +181,19 @@ void MemoryManager::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_)
     rasterizer = rasterizer_;
 }
 
-GPUVAddr MemoryManager::Map(GPUVAddr gpu_addr, DAddr dev_addr, std::size_t size, PTEKind kind, bool is_big_pages) {
-    if (is_big_pages)
-        return BigPageTableOp(gpu_addr, dev_addr, size, kind, EntryType::Mapped);
-    return PageTableOp(gpu_addr, dev_addr, size, kind, EntryType::Mapped);
+GPUVAddr MemoryManager::Map(GPUVAddr gpu_addr, DAddr dev_addr, std::size_t size, PTEKind kind,
+                            bool is_big_pages) {
+    if (is_big_pages) [[likely]] {
+        return BigPageTableOp<EntryType::Mapped>(gpu_addr, dev_addr, size, kind);
+    }
+    return PageTableOp<EntryType::Mapped>(gpu_addr, dev_addr, size, kind);
 }
 
 GPUVAddr MemoryManager::MapSparse(GPUVAddr gpu_addr, std::size_t size, bool is_big_pages) {
-    if (is_big_pages)
-        return BigPageTableOp(gpu_addr, 0, size, PTEKind::INVALID, EntryType::Reserved);
-    return PageTableOp(gpu_addr, 0, size, PTEKind::INVALID, EntryType::Reserved);
+    if (is_big_pages) [[likely]] {
+        return BigPageTableOp<EntryType::Reserved>(gpu_addr, 0, size, PTEKind::INVALID);
+    }
+    return PageTableOp<EntryType::Reserved>(gpu_addr, 0, size, PTEKind::INVALID);
 }
 
 void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
@@ -198,21 +207,26 @@ void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
     }
     page_stash.clear();
 
-    BigPageTableOp(gpu_addr, 0, size, PTEKind::INVALID, EntryType::Free);
-    PageTableOp(gpu_addr, 0, size, PTEKind::INVALID, EntryType::Free);
+    BigPageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
+    PageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
 }
 
 std::optional<DAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const {
     if (!IsWithinGPUAddressRange(gpu_addr)) [[unlikely]] {
         return std::nullopt;
     }
-    if (GetEntry(gpu_addr, true) != EntryType::Mapped) [[unlikely]] {
-        if (GetEntry(gpu_addr, false) != EntryType::Mapped)
+    if (GetEntry<true>(gpu_addr) != EntryType::Mapped) [[unlikely]] {
+        if (GetEntry<false>(gpu_addr) != EntryType::Mapped) {
             return std::nullopt;
-        const DAddr dev_addr_base = DAddr(page_table[PageEntryIndex(gpu_addr, false)]) << cpu_page_bits;
+        }
+
+        const DAddr dev_addr_base = static_cast<DAddr>(page_table[PageEntryIndex<false>(gpu_addr)])
+                                    << cpu_page_bits;
         return dev_addr_base + (gpu_addr & page_mask);
     }
-    const DAddr dev_addr_base = DAddr(big_page_table_dev[PageEntryIndex(gpu_addr, true)]) << cpu_page_bits;
+
+    const DAddr dev_addr_base =
+        static_cast<DAddr>(big_page_table_dev[PageEntryIndex<true>(gpu_addr)]) << cpu_page_bits;
     return dev_addr_base + (gpu_addr & big_page_mask);
 }
 
@@ -285,8 +299,10 @@ const u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) const {
 #pragma inline_recursion(on)
 #endif
 
-template <typename FuncMapped, typename FuncReserved, typename FuncUnmapped>
-inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size, bool is_big_page, FuncMapped&& func_mapped, FuncReserved&& func_reserved, FuncUnmapped&& func_unmapped) const {
+template <bool is_big_pages, typename FuncMapped, typename FuncReserved, typename FuncUnmapped>
+inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size,
+                                           FuncMapped&& func_mapped, FuncReserved&& func_reserved,
+                                           FuncUnmapped&& func_unmapped) const {
     using FuncMappedReturn =
         typename std::invoke_result<FuncMapped, std::size_t, std::size_t, std::size_t>::type;
     using FuncReservedReturn =
@@ -299,7 +315,7 @@ inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t si
     u64 used_page_size;
     u64 used_page_mask;
     u64 used_page_bits;
-    if (is_big_page) {
+    if constexpr (is_big_pages) {
         used_page_size = big_page_size;
         used_page_mask = big_page_mask;
         used_page_bits = big_page_bits;
@@ -316,7 +332,7 @@ inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t si
     while (remaining_size > 0) {
         const std::size_t copy_amount{
             (std::min)(static_cast<std::size_t>(used_page_size) - page_offset, remaining_size)};
-        auto entry = GetEntry(current_address, is_big_page);
+        auto entry = GetEntry<is_big_pages>(current_address);
         if (entry == EntryType::Mapped) [[likely]] {
             if constexpr (BOOL_BREAK_MAPPED) {
                 if (func_mapped(page_index, page_offset, copy_amount)) {
@@ -351,14 +367,18 @@ inline void MemoryManager::MemoryOperation(GPUVAddr gpu_src_addr, std::size_t si
     }
 }
 
-void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size, [[maybe_unused]] VideoCommon::CacheType which, bool unsafe) const {
-    auto set_to_zero = [&]([[maybe_unused]] std::size_t page_index, [[maybe_unused]] std::size_t offset, std::size_t copy_amount) {
+template <bool is_safe>
+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,
+                           [[maybe_unused]] std::size_t offset, std::size_t copy_amount) {
         std::memset(dest_buffer, 0, copy_amount);
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
     };
     auto mapped_normal = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const DAddr dev_addr_base = (DAddr(page_table[page_index]) << cpu_page_bits) + offset;
-        if (!unsafe) {
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+        if constexpr (is_safe) {
             rasterizer->FlushRegion(dev_addr_base, copy_amount, which);
         }
         u8* physical = memory.GetPointer<u8>(dev_addr_base);
@@ -366,8 +386,9 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
         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) {
-        const DAddr dev_addr_base = (DAddr(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
-        if (!unsafe) {
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
+        if constexpr (is_safe) {
             rasterizer->FlushRegion(dev_addr_base, copy_amount, which);
         }
         if (!IsBigPageContinuous(page_index)) [[unlikely]] {
@@ -378,28 +399,35 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
         }
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
     };
-    auto read_short_pages = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
+    auto read_short_pages = [&](std::size_t page_index, std::size_t offset,
+                                std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, mapped_normal, set_to_zero, set_to_zero);
+        MemoryOperation<false>(base, copy_amount, mapped_normal, set_to_zero, set_to_zero);
     };
-    MemoryOperation(gpu_src_addr, size, true, mapped_big, set_to_zero, read_short_pages);
+    MemoryOperation<true>(gpu_src_addr, size, mapped_big, set_to_zero, read_short_pages);
 }
 
-void MemoryManager::ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size, VideoCommon::CacheType which) const {
-    ReadBlockImpl(gpu_src_addr, dest_buffer, size, which, false);
+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);
 }
 
-void MemoryManager::ReadBlockUnsafe(GPUVAddr gpu_src_addr, void* dest_buffer, const std::size_t size) const {
-    ReadBlockImpl(gpu_src_addr, dest_buffer, size, VideoCommon::CacheType::None, true);
+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);
 }
 
-void MemoryManager::WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size, [[maybe_unused]] VideoCommon::CacheType which, bool unsafe) {
-    auto just_advance = [&]([[maybe_unused]] std::size_t page_index, [[maybe_unused]] std::size_t offset, std::size_t copy_amount) {
+template <bool is_safe>
+void MemoryManager::WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size,
+                                   [[maybe_unused]] VideoCommon::CacheType which) {
+    auto just_advance = [&]([[maybe_unused]] std::size_t page_index,
+                            [[maybe_unused]] std::size_t offset, std::size_t copy_amount) {
         src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
     };
     auto mapped_normal = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const DAddr dev_addr_base = (DAddr(page_table[page_index]) << cpu_page_bits) + offset;
-        if (!unsafe) {
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+        if constexpr (is_safe) {
             rasterizer->InvalidateRegion(dev_addr_base, copy_amount, which);
         }
         u8* physical = memory.GetPointer<u8>(dev_addr_base);
@@ -407,8 +435,9 @@ void MemoryManager::WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffe
         src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
     };
     auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const DAddr dev_addr_base = (DAddr(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
-        if (!unsafe) {
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
+        if constexpr (is_safe) {
             rasterizer->InvalidateRegion(dev_addr_base, copy_amount, which);
         }
         if (!IsBigPageContinuous(page_index)) [[unlikely]] {
@@ -419,23 +448,26 @@ void MemoryManager::WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffe
         }
         src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
     };
-    auto write_short_pages = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
+    auto write_short_pages = [&](std::size_t page_index, std::size_t offset,
+                                 std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, mapped_normal, just_advance, just_advance);
+        MemoryOperation<false>(base, copy_amount, mapped_normal, just_advance, just_advance);
     };
-    MemoryOperation(gpu_dest_addr, size, true, mapped_big, just_advance, write_short_pages);
+    MemoryOperation<true>(gpu_dest_addr, size, mapped_big, just_advance, write_short_pages);
 }
 
-void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size, VideoCommon::CacheType which) {
-    WriteBlockImpl(gpu_dest_addr, src_buffer, size, which, false);
+void MemoryManager::WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size,
+                               VideoCommon::CacheType which) {
+    WriteBlockImpl<true>(gpu_dest_addr, src_buffer, size, which);
 }
 
-void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size) {
-    WriteBlockImpl(gpu_dest_addr, src_buffer, size, VideoCommon::CacheType::None, true);
+void MemoryManager::WriteBlockUnsafe(GPUVAddr gpu_dest_addr, const void* src_buffer,
+                                     std::size_t size) {
+    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(gpu_dest_addr, src_buffer, size, VideoCommon::CacheType::None, true);
+    WriteBlockImpl<false>(gpu_dest_addr, src_buffer, size, VideoCommon::CacheType::None);
     accumulator.Add(gpu_dest_addr, size);
 }
 
@@ -446,18 +478,21 @@ void MemoryManager::FlushRegion(GPUVAddr gpu_addr, size_t size,
                           [[maybe_unused]] std::size_t copy_amount) {};
 
     auto mapped_normal = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const DAddr dev_addr_base = (DAddr(page_table[page_index]) << cpu_page_bits) + offset;
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
         rasterizer->FlushRegion(dev_addr_base, copy_amount, which);
     };
     auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const DAddr dev_addr_base = (DAddr(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
         rasterizer->FlushRegion(dev_addr_base, copy_amount, which);
     };
-    auto flush_short_pages = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
+    auto flush_short_pages = [&](std::size_t page_index, std::size_t offset,
+                                 std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, mapped_normal, do_nothing, do_nothing);
+        MemoryOperation<false>(base, copy_amount, mapped_normal, do_nothing, do_nothing);
     };
-    MemoryOperation(gpu_addr, size, true, mapped_big, do_nothing, flush_short_pages);
+    MemoryOperation<true>(gpu_addr, size, mapped_big, do_nothing, flush_short_pages);
 }
 
 bool MemoryManager::IsMemoryDirty(GPUVAddr gpu_addr, size_t size,
@@ -482,10 +517,10 @@ bool MemoryManager::IsMemoryDirty(GPUVAddr gpu_addr, size_t size,
     auto check_short_pages = [&](std::size_t page_index, std::size_t offset,
                                  std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, mapped_normal, do_nothing, do_nothing);
+        MemoryOperation<false>(base, copy_amount, mapped_normal, do_nothing, do_nothing);
         return result;
     };
-    MemoryOperation(gpu_addr, size, true, mapped_big, do_nothing, check_short_pages);
+    MemoryOperation<true>(gpu_addr, size, mapped_big, do_nothing, check_short_pages);
     return result;
 }
 
@@ -522,10 +557,10 @@ size_t MemoryManager::MaxContinuousRange(GPUVAddr gpu_addr, size_t size) const {
     auto check_short_pages = [&](std::size_t page_index, std::size_t offset,
                                  std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, short_check, fail, fail);
+        MemoryOperation<false>(base, copy_amount, short_check, fail, fail);
         return result;
     };
-    MemoryOperation(gpu_addr, size, true, big_check, fail, check_short_pages);
+    MemoryOperation<true>(gpu_addr, size, big_check, fail, check_short_pages);
     return range_so_far;
 }
 
@@ -541,18 +576,21 @@ void MemoryManager::InvalidateRegion(GPUVAddr gpu_addr, size_t size,
                           [[maybe_unused]] std::size_t copy_amount) {};
 
     auto mapped_normal = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const DAddr dev_addr_base = (DAddr(page_table[page_index]) << cpu_page_bits) + offset;
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
         rasterizer->InvalidateRegion(dev_addr_base, copy_amount, which);
     };
     auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const DAddr dev_addr_base = (DAddr(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
+        const DAddr dev_addr_base =
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
         rasterizer->InvalidateRegion(dev_addr_base, copy_amount, which);
     };
-    auto invalidate_short_pages = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
+    auto invalidate_short_pages = [&](std::size_t page_index, std::size_t offset,
+                                      std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, mapped_normal, do_nothing, do_nothing);
+        MemoryOperation<false>(base, copy_amount, mapped_normal, do_nothing, do_nothing);
     };
-    MemoryOperation(gpu_addr, size, true, mapped_big, do_nothing, invalidate_short_pages);
+    MemoryOperation<true>(gpu_addr, size, mapped_big, do_nothing, invalidate_short_pages);
 }
 
 void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size,
@@ -564,7 +602,7 @@ void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std
 }
 
 bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const {
-    if (GetEntry(gpu_addr, true) == EntryType::Mapped) [[likely]] {
+    if (GetEntry<true>(gpu_addr) == EntryType::Mapped) [[likely]] {
         size_t page_index = gpu_addr >> big_page_bits;
         if (IsBigPageContinuous(page_index)) [[likely]] {
             const std::size_t page{(page_index & big_page_mask) + size};
@@ -573,7 +611,7 @@ bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const {
         const std::size_t page{(gpu_addr & Core::DEVICE_PAGEMASK) + size};
         return page <= Core::DEVICE_PAGESIZE;
     }
-    if (GetEntry(gpu_addr, false) != EntryType::Mapped) {
+    if (GetEntry<false>(gpu_addr) != EntryType::Mapped) {
         return false;
     }
     const std::size_t page{(gpu_addr & Core::DEVICE_PAGEMASK) + size};
@@ -611,10 +649,10 @@ bool MemoryManager::IsContinuousRange(GPUVAddr gpu_addr, std::size_t size) const
     auto check_short_pages = [&](std::size_t page_index, std::size_t offset,
                                  std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, short_check, fail, fail);
+        MemoryOperation<false>(base, copy_amount, short_check, fail, fail);
         return !result;
     };
-    MemoryOperation(gpu_addr, size, true, big_check, fail, check_short_pages);
+    MemoryOperation<true>(gpu_addr, size, big_check, fail, check_short_pages);
     return result;
 }
 
@@ -627,12 +665,13 @@ bool MemoryManager::IsFullyMappedRange(GPUVAddr gpu_addr, std::size_t size) cons
     };
     auto pass = [&]([[maybe_unused]] std::size_t page_index, [[maybe_unused]] std::size_t offset,
                     [[maybe_unused]] std::size_t copy_amount) { return false; };
-    auto check_short_pages = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
+    auto check_short_pages = [&](std::size_t page_index, std::size_t offset,
+                                 std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, pass, pass, fail);
+        MemoryOperation<false>(base, copy_amount, pass, pass, fail);
         return !result;
     };
-    MemoryOperation(gpu_addr, size, true, pass, fail, check_short_pages);
+    MemoryOperation<true>(gpu_addr, size, pass, fail, check_short_pages);
     return result;
 }
 
@@ -644,9 +683,13 @@ MemoryManager::GetSubmappedRange(GPUVAddr gpu_addr, std::size_t size) const {
 }
 
 template <bool is_gpu_address>
-void MemoryManager::GetSubmappedRangeImpl(GPUVAddr gpu_addr, std::size_t size, boost::container::small_vector<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>, 32>& result)
+void MemoryManager::GetSubmappedRangeImpl(
+    GPUVAddr gpu_addr, std::size_t size,
+    boost::container::small_vector<
+        std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>, 32>& result)
     const {
-    std::optional<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>> last_segment{};
+    std::optional<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>>
+        last_segment{};
     std::optional<DAddr> old_page_addr{};
     const auto split = [&last_segment, &result]([[maybe_unused]] std::size_t page_index,
                                                 [[maybe_unused]] std::size_t offset,
@@ -702,9 +745,9 @@ void MemoryManager::GetSubmappedRangeImpl(GPUVAddr gpu_addr, std::size_t size, b
     };
     auto do_short_pages = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
         GPUVAddr base = (page_index << big_page_bits) + offset;
-        MemoryOperation(base, copy_amount, false, extend_size_short, split, split);
+        MemoryOperation<false>(base, copy_amount, extend_size_short, split, split);
     };
-    MemoryOperation(gpu_addr, size, true, extend_size_big, split, do_short_pages);
+    MemoryOperation<true>(gpu_addr, size, extend_size_big, split, do_short_pages);
     split(0, 0, 0);
 }
 

src/video_core/memory_manager.h

@@ -45,7 +45,7 @@ public:
 
     static constexpr bool HAS_FLUSH_INVALIDATION = true;
 
-    inline size_t GetID() const noexcept {
+    size_t GetID() const {
         return unique_identifier;
     }
 
@@ -66,15 +66,16 @@ public:
     [[nodiscard]] const u8* GetPointer(GPUVAddr addr) const;
 
     template <typename T>
-    [[nodiscard]] inline T* GetPointer(GPUVAddr addr) noexcept {
-        const auto address = GpuToCpuAddress(addr);
-        if (!address)
+    [[nodiscard]] T* GetPointer(GPUVAddr addr) {
+        const auto address{GpuToCpuAddress(addr)};
+        if (!address) {
             return {};
+        }
         return memory.GetPointer<T>(*address);
     }
 
     template <typename T>
-    [[nodiscard]] inline const T* GetPointer(GPUVAddr addr) const noexcept {
+    [[nodiscard]] const T* GetPointer(GPUVAddr addr) const {
         return GetPointer<T*>(addr);
     }
 
@@ -84,9 +85,12 @@ public:
      * in the Host Memory counterpart. Note: This functions cause Host GPU Memory
      * Flushes and Invalidations, respectively to each operation.
      */
-    void ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size, VideoCommon::CacheType which = VideoCommon::CacheType::All) const;
-    void WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size, VideoCommon::CacheType which = VideoCommon::CacheType::All);
-    void CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size, VideoCommon::CacheType which = VideoCommon::CacheType::All);
+    void ReadBlock(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
+                   VideoCommon::CacheType which = VideoCommon::CacheType::All) const;
+    void WriteBlock(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size,
+                    VideoCommon::CacheType which = VideoCommon::CacheType::All);
+    void CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size,
+                   VideoCommon::CacheType which = VideoCommon::CacheType::All);
 
     /**
      * ReadBlockUnsafe and WriteBlockUnsafe are special versions of ReadBlock and
@@ -156,14 +160,21 @@ public:
     u8* GetSpan(const GPUVAddr src_addr, const std::size_t size);
 
 private:
-    template <typename FuncMapped, typename FuncReserved, typename FuncUnmapped>
-    inline void MemoryOperation(GPUVAddr gpu_src_addr, std::size_t size, bool is_big_page, FuncMapped&& func_mapped, FuncReserved&& func_reserved, FuncUnmapped&& func_unmapped) const;
+    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;
 
-    void ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size, VideoCommon::CacheType which, bool unsafe) const;
-    void WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size, VideoCommon::CacheType which, bool unsafe);
+    template <bool is_safe>
+    void ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std::size_t size,
+                       VideoCommon::CacheType which) const;
 
-    [[nodiscard]] std::size_t PageEntryIndex(GPUVAddr gpu_addr, bool is_big_page) const {
-        if (is_big_page) {
+    template <bool is_safe>
+    void WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffer, std::size_t size,
+                        VideoCommon::CacheType which);
+
+    template <bool is_big_page>
+    [[nodiscard]] std::size_t PageEntryIndex(GPUVAddr gpu_addr) const {
+        if constexpr (is_big_page) {
             return (gpu_addr >> big_page_bits) & big_page_table_mask;
         } else {
             return (gpu_addr >> page_bits) & page_table_mask;
@@ -176,7 +187,9 @@ private:
     template <bool is_gpu_address>
     void GetSubmappedRangeImpl(
         GPUVAddr gpu_addr, std::size_t size,
-        boost::container::small_vector<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>, 32>& result) const;
+        boost::container::small_vector<
+            std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>, 32>&
+            result) const;
 
     Core::System& system;
     MaxwellDeviceMemoryManager& memory;
@@ -206,11 +219,19 @@ private:
     std::vector<u64> entries;
     std::vector<u64> big_entries;
 
-    GPUVAddr PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size, PTEKind kind, EntryType entry_type);
-    GPUVAddr BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size, PTEKind kind, EntryType entry_type);
+    template <EntryType entry_type>
+    GPUVAddr PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size,
+                         PTEKind kind);
 
-    inline EntryType GetEntry(size_t position, bool is_big_page) const;
-    inline void SetEntry(size_t position, EntryType entry, bool is_big_page);
+    template <EntryType entry_type>
+    GPUVAddr BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size,
+                            PTEKind kind);
+
+    template <bool is_big_page>
+    inline EntryType GetEntry(size_t position) const;
+
+    template <bool is_big_page>
+    inline void SetEntry(size_t position, EntryType entry);
 
     Common::MultiLevelPageTable<u32> page_table;
     Common::RangeMap<GPUVAddr, PTEKind> kind_map;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -485,7 +485,6 @@ void RasterizerOpenGL::FlushRegion(DAddr addr, u64 size, VideoCommon::CacheType
         texture_cache.DownloadMemory(addr, size);
     }
     if ((True(which & VideoCommon::CacheType::BufferCache))) {
-        std::scoped_lock lock{buffer_cache.mutex};
         buffer_cache.DownloadMemory(addr, size);
     }
     if ((True(which & VideoCommon::CacheType::QueryCache))) {

src/video_core/renderer_vulkan/vk_command_pool.cpp

@@ -1,9 +1,13 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
 #include <cstddef>
 
 #include "video_core/renderer_vulkan/vk_command_pool.h"
+#include "video_core/renderer_vulkan/vk_master_semaphore.h"
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
@@ -14,32 +18,52 @@ constexpr size_t COMMAND_BUFFER_POOL_SIZE = 4;
 struct CommandPool::Pool {
     vk::CommandPool handle;
     vk::CommandBuffers cmdbufs;
+    u64 tick;
 };
 
 CommandPool::CommandPool(MasterSemaphore& master_semaphore_, const Device& device_)
-    : ResourcePool(master_semaphore_, COMMAND_BUFFER_POOL_SIZE), device{device_} {}
+    : master_semaphore{master_semaphore_}, device{device_} {}
 
 CommandPool::~CommandPool() = default;
 
-void CommandPool::Allocate(size_t begin, size_t end) {
-    // Command buffers are going to be committed, recorded, executed every single usage cycle.
-    // They are also going to be reset when committed.
+void CommandPool::AllocatePool() {
     Pool& pool = pools.emplace_back();
     pool.handle = device.GetLogical().CreateCommandPool({
         .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,
+        .flags = VK_COMMAND_POOL_CREATE_TRANSIENT_BIT,
         .queueFamilyIndex = device.GetGraphicsFamily(),
     });
     pool.cmdbufs = pool.handle.Allocate(COMMAND_BUFFER_POOL_SIZE);
+    pool.tick = 0;
+}
+
+void CommandPool::AcquirePool() {
+    if (!pools.empty()) {
+        master_semaphore.Refresh();
+        const u64 gpu_tick = master_semaphore.KnownGpuTick();
+        for (size_t i = 0; i < pools.size(); ++i) {
+            const size_t candidate = (current_pool + 1 + i) % pools.size();
+            if (gpu_tick >= pools[candidate].tick) {
+                current_pool = candidate;
+                current_index = 0;
+                pools[current_pool].handle.Reset();
+                return;
+            }
+        }
+    }
+    AllocatePool();
+    current_pool = pools.size() - 1;
+    current_index = 0;
 }
 
 VkCommandBuffer CommandPool::Commit() {
-    const size_t index = CommitResource();
-    const auto pool_index = index / COMMAND_BUFFER_POOL_SIZE;
-    const auto sub_index = index % COMMAND_BUFFER_POOL_SIZE;
-    return pools[pool_index].cmdbufs[sub_index];
+    if (pools.empty() || current_index >= COMMAND_BUFFER_POOL_SIZE) {
+        AcquirePool();
+    }
+    Pool& pool = pools[current_pool];
+    pool.tick = master_semaphore.CurrentTick();
+    return pool.cmdbufs[current_index++];
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_command_pool.h

@@ -1,3 +1,6 @@
+// SPDX-FileCopyrightText: Copyright 2026 Eden Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
 // SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
@@ -6,7 +9,7 @@
 #include <cstddef>
 #include <vector>
 
-#include "video_core/renderer_vulkan/vk_resource_pool.h"
+#include "common/common_types.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
 namespace Vulkan {
@@ -14,20 +17,24 @@ namespace Vulkan {
 class Device;
 class MasterSemaphore;
 
-class CommandPool final : public ResourcePool {
+class CommandPool final {
 public:
     explicit CommandPool(MasterSemaphore& master_semaphore_, const Device& device_);
-    ~CommandPool() override;
-
-    void Allocate(size_t begin, size_t end) override;
+    ~CommandPool();
 
     VkCommandBuffer Commit();
 
 private:
     struct Pool;
 
+    void AllocatePool();
+    void AcquirePool();
+
+    MasterSemaphore& master_semaphore;
     const Device& device;
     std::vector<Pool> pools;
+    size_t current_pool = 0;
+    size_t current_index = 0;
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -4,6 +4,7 @@
 // SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#include <algorithm>
 #include <array>
 #include <memory>
 #include <numeric>
@@ -17,6 +18,8 @@
 #include "common/div_ceil.h"
 #include "common/vector_math.h"
 #include "video_core/host_shaders/astc_decoder_comp_spv.h"
+#include "video_core/host_shaders/astc_decoder_frag_spv.h"
+#include "video_core/host_shaders/astc_decoder_vert_spv.h"
 #include "video_core/host_shaders/queries_prefix_scan_sum_comp_spv.h"
 #include "video_core/host_shaders/queries_prefix_scan_sum_nosubgroups_comp_spv.h"
 #include "video_core/host_shaders/resolve_conditional_render_comp_spv.h"
@@ -25,8 +28,11 @@
 #include "video_core/host_shaders/block_linear_unswizzle_3d_bcn_comp_spv.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
 #include "video_core/surface.h"
+#include "video_core/renderer_vulkan/maxwell_to_vk.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_render_pass_cache.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_shader_util.h"
 #include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
 #include "video_core/texture_cache/accelerated_swizzle.h"
@@ -613,7 +619,394 @@ void ASTCDecoderPass::Assemble(Image& image, const StagingBufferRef& map,
         cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT,
                        vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, 0, image_barrier);
     });
-    scheduler.Finish();
+}
+
+namespace {
+struct AstcFragPushConstants {
+    std::array<u32, 2> blocks_dims;
+    u32 layer_stride;
+    u32 block_size;
+    u32 x_shift;
+    u32 block_height;
+    u32 block_height_mask;
+    u32 dest_layer;
+};
+
+constexpr VkDescriptorSetLayoutBinding ASTC_FRAG_BINDING{
+    .binding = 0,
+    .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+    .descriptorCount = 1,
+    .stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT,
+    .pImmutableSamplers = nullptr,
+};
+
+constexpr VkDescriptorUpdateTemplateEntry ASTC_FRAG_TEMPLATE{
+    .dstBinding = 0,
+    .dstArrayElement = 0,
+    .descriptorCount = 1,
+    .descriptorType = VK_DESCRIPTOR_TYPE_STORAGE_BUFFER,
+    .offset = 0,
+    .stride = sizeof(DescriptorUpdateEntry),
+};
+
+constexpr DescriptorBankInfo ASTC_FRAG_BANK_INFO{
+    .uniform_buffers = 0,
+    .storage_buffers = 1,
+    .texture_buffers = 0,
+    .image_buffers = 0,
+    .textures = 0,
+    .images = 0,
+    .score = 1,
+};
+
+constexpr VkPushConstantRange ASTC_FRAG_PUSH_CONSTANT_RANGE{
+    .stageFlags = VK_SHADER_STAGE_FRAGMENT_BIT,
+    .offset = 0,
+    .size = static_cast<u32>(sizeof(AstcFragPushConstants)),
+};
+
+constexpr VkPipelineVertexInputStateCreateInfo ASTC_FRAG_VERTEX_INPUT{
+    .sType = VK_STRUCTURE_TYPE_PIPELINE_VERTEX_INPUT_STATE_CREATE_INFO,
+    .pNext = nullptr,
+    .flags = 0,
+    .vertexBindingDescriptionCount = 0,
+    .pVertexBindingDescriptions = nullptr,
+    .vertexAttributeDescriptionCount = 0,
+    .pVertexAttributeDescriptions = nullptr,
+};
+
+constexpr VkPipelineInputAssemblyStateCreateInfo ASTC_FRAG_INPUT_ASSEMBLY{
+    .sType = VK_STRUCTURE_TYPE_PIPELINE_INPUT_ASSEMBLY_STATE_CREATE_INFO,
+    .pNext = nullptr,
+    .flags = 0,
+    .topology = VK_PRIMITIVE_TOPOLOGY_TRIANGLE_LIST,
+    .primitiveRestartEnable = VK_FALSE,
+};
+
+constexpr VkPipelineViewportStateCreateInfo ASTC_FRAG_VIEWPORT{
+    .sType = VK_STRUCTURE_TYPE_PIPELINE_VIEWPORT_STATE_CREATE_INFO,
+    .pNext = nullptr,
+    .flags = 0,
+    .viewportCount = 1,
+    .pViewports = nullptr,
+    .scissorCount = 1,
+    .pScissors = nullptr,
+};
+
+constexpr VkPipelineRasterizationStateCreateInfo ASTC_FRAG_RASTERIZATION{
+    .sType = VK_STRUCTURE_TYPE_PIPELINE_RASTERIZATION_STATE_CREATE_INFO,
+    .pNext = nullptr,
+    .flags = 0,
+    .depthClampEnable = VK_FALSE,
+    .rasterizerDiscardEnable = VK_FALSE,
+    .polygonMode = VK_POLYGON_MODE_FILL,
+    .cullMode = VK_CULL_MODE_NONE,
+    .frontFace = VK_FRONT_FACE_COUNTER_CLOCKWISE,
+    .depthBiasEnable = VK_FALSE,
+    .depthBiasConstantFactor = 0.0f,
+    .depthBiasClamp = 0.0f,
+    .depthBiasSlopeFactor = 0.0f,
+    .lineWidth = 1.0f,
+};
+
+constexpr VkPipelineMultisampleStateCreateInfo ASTC_FRAG_MULTISAMPLE{
+    .sType = VK_STRUCTURE_TYPE_PIPELINE_MULTISAMPLE_STATE_CREATE_INFO,
+    .pNext = nullptr,
+    .flags = 0,
+    .rasterizationSamples = VK_SAMPLE_COUNT_1_BIT,
+    .sampleShadingEnable = VK_FALSE,
+    .minSampleShading = 0.0f,
+    .pSampleMask = nullptr,
+    .alphaToCoverageEnable = VK_FALSE,
+    .alphaToOneEnable = VK_FALSE,
+};
+
+constexpr VkPipelineColorBlendAttachmentState ASTC_FRAG_BLEND_ATTACHMENT{
+    .blendEnable = VK_FALSE,
+    .srcColorBlendFactor = VK_BLEND_FACTOR_ONE,
+    .dstColorBlendFactor = VK_BLEND_FACTOR_ZERO,
+    .colorBlendOp = VK_BLEND_OP_ADD,
+    .srcAlphaBlendFactor = VK_BLEND_FACTOR_ONE,
+    .dstAlphaBlendFactor = VK_BLEND_FACTOR_ZERO,
+    .alphaBlendOp = VK_BLEND_OP_ADD,
+    .colorWriteMask = VK_COLOR_COMPONENT_R_BIT | VK_COLOR_COMPONENT_G_BIT |
+                      VK_COLOR_COMPONENT_B_BIT | VK_COLOR_COMPONENT_A_BIT,
+};
+
+constexpr std::array ASTC_FRAG_DYNAMIC_STATES{VK_DYNAMIC_STATE_VIEWPORT,
+                                              VK_DYNAMIC_STATE_SCISSOR};
+} // Anonymous namespace
+
+ASTCDecoderFragmentPass::ASTCDecoderFragmentPass(
+    const Device& device_, Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue_,
+    RenderPassCache& render_pass_cache_)
+    : device{device_}, scheduler{scheduler_},
+      compute_pass_descriptor_queue{compute_pass_descriptor_queue_},
+      render_pass_cache{render_pass_cache_},
+      descriptor_set_layout{device.GetLogical().CreateDescriptorSetLayout({
+          .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_SET_LAYOUT_CREATE_INFO,
+          .pNext = nullptr,
+          .flags = 0,
+          .bindingCount = 1,
+          .pBindings = &ASTC_FRAG_BINDING,
+      })},
+      descriptor_template{device.GetLogical().CreateDescriptorUpdateTemplate({
+          .sType = VK_STRUCTURE_TYPE_DESCRIPTOR_UPDATE_TEMPLATE_CREATE_INFO,
+          .pNext = nullptr,
+          .flags = 0,
+          .descriptorUpdateEntryCount = 1,
+          .pDescriptorUpdateEntries = &ASTC_FRAG_TEMPLATE,
+          .templateType = VK_DESCRIPTOR_UPDATE_TEMPLATE_TYPE_DESCRIPTOR_SET,
+          .descriptorSetLayout = *descriptor_set_layout,
+          .pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS,
+          .pipelineLayout = VK_NULL_HANDLE,
+          .set = 0,
+      })},
+      pipeline_layout{device.GetLogical().CreatePipelineLayout({
+          .sType = VK_STRUCTURE_TYPE_PIPELINE_LAYOUT_CREATE_INFO,
+          .pNext = nullptr,
+          .flags = 0,
+          .setLayoutCount = 1,
+          .pSetLayouts = descriptor_set_layout.address(),
+          .pushConstantRangeCount = 1,
+          .pPushConstantRanges = &ASTC_FRAG_PUSH_CONSTANT_RANGE,
+      })},
+      descriptor_allocator{descriptor_pool_.Allocator(device_, scheduler_, *descriptor_set_layout,
+                                                       ASTC_FRAG_BANK_INFO)},
+      vertex_shader{BuildShader(device, ASTC_DECODER_VERT_SPV)},
+      fragment_shader{BuildShader(device, ASTC_DECODER_FRAG_SPV)} {}
+
+ASTCDecoderFragmentPass::~ASTCDecoderFragmentPass() = default;
+
+VkPipeline ASTCDecoderFragmentPass::FindOrEmplacePipeline(VkRenderPass render_pass) {
+    const auto it = std::ranges::find(pipeline_keys, render_pass);
+    if (it != pipeline_keys.end()) {
+        return *pipelines[std::distance(pipeline_keys.begin(), it)];
+    }
+    pipeline_keys.push_back(render_pass);
+    const std::array stages{
+        VkPipelineShaderStageCreateInfo{
+            .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+            .pNext = nullptr,
+            .flags = 0,
+            .stage = VK_SHADER_STAGE_VERTEX_BIT,
+            .module = *vertex_shader,
+            .pName = "main",
+            .pSpecializationInfo = nullptr,
+        },
+        VkPipelineShaderStageCreateInfo{
+            .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+            .pNext = nullptr,
+            .flags = 0,
+            .stage = VK_SHADER_STAGE_FRAGMENT_BIT,
+            .module = *fragment_shader,
+            .pName = "main",
+            .pSpecializationInfo = nullptr,
+        },
+    };
+    const VkPipelineColorBlendStateCreateInfo color_blend{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_COLOR_BLEND_STATE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .logicOpEnable = VK_FALSE,
+        .logicOp = VK_LOGIC_OP_CLEAR,
+        .attachmentCount = 1,
+        .pAttachments = &ASTC_FRAG_BLEND_ATTACHMENT,
+        .blendConstants = {0.0f, 0.0f, 0.0f, 0.0f},
+    };
+    const VkPipelineDynamicStateCreateInfo dynamic_state{
+        .sType = VK_STRUCTURE_TYPE_PIPELINE_DYNAMIC_STATE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .dynamicStateCount = static_cast<u32>(ASTC_FRAG_DYNAMIC_STATES.size()),
+        .pDynamicStates = ASTC_FRAG_DYNAMIC_STATES.data(),
+    };
+    pipelines.push_back(device.GetLogical().CreateGraphicsPipeline({
+        .sType = VK_STRUCTURE_TYPE_GRAPHICS_PIPELINE_CREATE_INFO,
+        .pNext = nullptr,
+        .flags = 0,
+        .stageCount = static_cast<u32>(stages.size()),
+        .pStages = stages.data(),
+        .pVertexInputState = &ASTC_FRAG_VERTEX_INPUT,
+        .pInputAssemblyState = &ASTC_FRAG_INPUT_ASSEMBLY,
+        .pTessellationState = nullptr,
+        .pViewportState = &ASTC_FRAG_VIEWPORT,
+        .pRasterizationState = &ASTC_FRAG_RASTERIZATION,
+        .pMultisampleState = &ASTC_FRAG_MULTISAMPLE,
+        .pDepthStencilState = nullptr,
+        .pColorBlendState = &color_blend,
+        .pDynamicState = &dynamic_state,
+        .layout = *pipeline_layout,
+        .renderPass = render_pass,
+        .subpass = 0,
+        .basePipelineHandle = VK_NULL_HANDLE,
+        .basePipelineIndex = 0,
+    }));
+    return *pipelines.back();
+}
+
+void ASTCDecoderFragmentPass::Assemble(Image& image, const StagingBufferRef& map,
+                                       std::span<const VideoCommon::SwizzleParameters> swizzles,
+                                       VideoCore::Surface::PixelFormat decoded_format) {
+    using namespace VideoCommon::Accelerated;
+    while (!frame_resources.empty() && scheduler.IsFree(frame_resources.front().tick)) {
+        frame_resources.pop_front();
+    }
+    const std::array<u32, 2> block_dims{
+        VideoCore::Surface::DefaultBlockWidth(image.info.format),
+        VideoCore::Surface::DefaultBlockHeight(image.info.format),
+    };
+    RenderPassKey key{};
+    key.color_formats.fill(VideoCore::Surface::PixelFormat::Invalid);
+    key.color_formats[0] = decoded_format;
+    key.depth_format = VideoCore::Surface::PixelFormat::Invalid;
+    key.samples = VK_SAMPLE_COUNT_1_BIT;
+    const VkRenderPass render_pass = render_pass_cache.Get(key);
+    const VkPipeline pipeline = FindOrEmplacePipeline(render_pass);
+    const VkFormat vk_format =
+        MaxwellToVK::SurfaceFormat(device, FormatType::Optimal, false, decoded_format).format;
+    const VkImage vk_image = image.Handle();
+    const VkImageAspectFlags aspect_mask = image.AspectMask();
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+    const bool is_initialized = image.ExchangeInitialization();
+    scheduler.Record([vk_image, aspect_mask, is_initialized](vk::CommandBuffer cmdbuf) {
+        const VkImageMemoryBarrier barrier{
+            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = static_cast<VkAccessFlags>(
+                is_initialized ? VK_ACCESS_SHADER_READ_BIT : VK_ACCESS_NONE),
+            .dstAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
+            .oldLayout = is_initialized ? VK_IMAGE_LAYOUT_GENERAL : VK_IMAGE_LAYOUT_UNDEFINED,
+            .newLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .image = vk_image,
+            .subresourceRange{
+                .aspectMask = aspect_mask,
+                .baseMipLevel = 0,
+                .levelCount = VK_REMAINING_MIP_LEVELS,
+                .baseArrayLayer = 0,
+                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+            },
+        };
+        cmdbuf.PipelineBarrier(vk::PIPELINE_STAGE_GRAPHICS_COMPUTE_TRANSFER,
+                               VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT, 0, barrier);
+    });
+
+    for (const VideoCommon::SwizzleParameters& swizzle : swizzles) {
+        const size_t input_offset = swizzle.buffer_offset + map.offset;
+        const auto params = MakeBlockLinearSwizzle2DParams(swizzle, image.info);
+        const u32 level = swizzle.level;
+        const u32 width = std::max(1u, image.info.size.width >> level);
+        const u32 height = std::max(1u, image.info.size.height >> level);
+        const u32 layers = image.info.resources.layers;
+        for (u32 layer = 0; layer < layers; ++layer) {
+            vk::ImageView view = device.GetLogical().CreateImageView(VkImageViewCreateInfo{
+                .sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
+                .pNext = nullptr,
+                .flags = 0,
+                .image = vk_image,
+                .viewType = VK_IMAGE_VIEW_TYPE_2D,
+                .format = vk_format,
+                .components{},
+                .subresourceRange{
+                    .aspectMask = aspect_mask,
+                    .baseMipLevel = level,
+                    .levelCount = 1,
+                    .baseArrayLayer = layer,
+                    .layerCount = 1,
+                },
+            });
+            vk::Framebuffer framebuffer =
+                device.GetLogical().CreateFramebuffer(VkFramebufferCreateInfo{
+                    .sType = VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO,
+                    .pNext = nullptr,
+                    .flags = 0,
+                    .renderPass = render_pass,
+                    .attachmentCount = 1,
+                    .pAttachments = view.address(),
+                    .width = width,
+                    .height = height,
+                    .layers = 1,
+                });
+            const AstcFragPushConstants pc{
+                .blocks_dims = block_dims,
+                .layer_stride = params.layer_stride,
+                .block_size = params.block_size,
+                .x_shift = params.x_shift,
+                .block_height = params.block_height,
+                .block_height_mask = params.block_height_mask,
+                .dest_layer = layer,
+            };
+            compute_pass_descriptor_queue.Acquire(scheduler, 1);
+            compute_pass_descriptor_queue.AddBuffer(map.buffer, input_offset,
+                                                    image.guest_size_bytes - swizzle.buffer_offset);
+            const void* const descriptor_data{compute_pass_descriptor_queue.UpdateData()};
+            scheduler.Record([this, pipeline, render_pass, descriptor_data, pc, width, height,
+                              fb = *framebuffer](vk::CommandBuffer cmdbuf) {
+                const VkDescriptorSet set = descriptor_allocator.Commit();
+                device.GetLogical().UpdateDescriptorSet(set, *descriptor_template, descriptor_data);
+                const VkRenderPassBeginInfo begin{
+                    .sType = VK_STRUCTURE_TYPE_RENDER_PASS_BEGIN_INFO,
+                    .pNext = nullptr,
+                    .renderPass = render_pass,
+                    .framebuffer = fb,
+                    .renderArea{.offset = {0, 0}, .extent = {width, height}},
+                    .clearValueCount = 0,
+                    .pClearValues = nullptr,
+                };
+                const VkViewport viewport{
+                    .x = 0.0f,
+                    .y = 0.0f,
+                    .width = static_cast<float>(width),
+                    .height = static_cast<float>(height),
+                    .minDepth = 0.0f,
+                    .maxDepth = 1.0f,
+                };
+                const VkRect2D scissor{.offset = {0, 0}, .extent = {width, height}};
+                cmdbuf.BeginRenderPass(begin, VK_SUBPASS_CONTENTS_INLINE);
+                cmdbuf.BindPipeline(VK_PIPELINE_BIND_POINT_GRAPHICS, pipeline);
+                cmdbuf.SetViewport(0, viewport);
+                cmdbuf.SetScissor(0, scissor);
+                cmdbuf.BindDescriptorSets(VK_PIPELINE_BIND_POINT_GRAPHICS, *pipeline_layout, 0, set,
+                                          {});
+                cmdbuf.PushConstants(*pipeline_layout, VK_SHADER_STAGE_FRAGMENT_BIT, pc);
+                cmdbuf.Draw(3, 1, 0, 0);
+                cmdbuf.EndRenderPass();
+            });
+            frame_resources.push_back(FrameResources{
+                .tick = scheduler.CurrentTick(),
+                .view = std::move(view),
+                .framebuffer = std::move(framebuffer),
+            });
+        }
+    }
+
+    scheduler.Record([vk_image, aspect_mask](vk::CommandBuffer cmdbuf) {
+        const VkImageMemoryBarrier barrier{
+            .sType = VK_STRUCTURE_TYPE_IMAGE_MEMORY_BARRIER,
+            .pNext = nullptr,
+            .srcAccessMask = VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT,
+            .dstAccessMask = VK_ACCESS_SHADER_READ_BIT,
+            .oldLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL,
+            .newLayout = VK_IMAGE_LAYOUT_GENERAL,
+            .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+            .image = vk_image,
+            .subresourceRange{
+                .aspectMask = aspect_mask,
+                .baseMipLevel = 0,
+                .levelCount = VK_REMAINING_MIP_LEVELS,
+                .baseArrayLayer = 0,
+                .layerCount = VK_REMAINING_ARRAY_LAYERS,
+            },
+        };
+        cmdbuf.PipelineBarrier(VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT,
+                               vk::PIPELINE_STAGE_GRAPHICS_COMPUTE, 0, barrier);
+    });
+    scheduler.InvalidateState();
 }
 
 constexpr u32 BL3D_BINDING_INPUT_BUFFER  = 0;

src/video_core/renderer_vulkan/vk_compute_pass.h

@@ -6,12 +6,15 @@
 
 #pragma once
 
+#include <deque>
 #include <optional>
 #include <span>
 #include <utility>
+#include <vector>
 
 #include "common/common_types.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/surface.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
 #include "video_core/texture_cache/types.h"
@@ -137,6 +140,44 @@ private:
     MemoryAllocator& memory_allocator;
 };
 
+class RenderPassCache;
+
+class ASTCDecoderFragmentPass final {
+public:
+    explicit ASTCDecoderFragmentPass(const Device& device_, Scheduler& scheduler_,
+                                     DescriptorPool& descriptor_pool_,
+                                     ComputePassDescriptorQueue& compute_pass_descriptor_queue_,
+                                     RenderPassCache& render_pass_cache_);
+    ~ASTCDecoderFragmentPass();
+
+    void Assemble(Image& image, const StagingBufferRef& map,
+                  std::span<const VideoCommon::SwizzleParameters> swizzles,
+                  VideoCore::Surface::PixelFormat decoded_format);
+
+private:
+    VkPipeline FindOrEmplacePipeline(VkRenderPass render_pass);
+
+    const Device& device;
+    Scheduler& scheduler;
+    ComputePassDescriptorQueue& compute_pass_descriptor_queue;
+    RenderPassCache& render_pass_cache;
+    vk::DescriptorSetLayout descriptor_set_layout;
+    vk::DescriptorUpdateTemplate descriptor_template;
+    vk::PipelineLayout pipeline_layout;
+    DescriptorAllocator descriptor_allocator;
+    vk::ShaderModule vertex_shader;
+    vk::ShaderModule fragment_shader;
+    std::vector<VkRenderPass> pipeline_keys;
+    std::vector<vk::Pipeline> pipelines;
+
+    struct FrameResources {
+        u64 tick;
+        vk::ImageView view;
+        vk::Framebuffer framebuffer;
+    };
+    std::deque<FrameResources> frame_resources;
+};
+
 class BlockLinearUnswizzle3DPass final : public ComputePass {
 public:
     explicit BlockLinearUnswizzle3DPass(const Device& device_, Scheduler& scheduler_,

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -167,6 +167,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
     }
     const Shader::Stage stage{program.stage};
     const bool has_geometry{key.unique_hashes[4] != 0 && !programs[4].is_geometry_passthrough};
+    const bool has_tessellation{key.unique_hashes[3] != 0};
     const bool gl_ndc{key.state.ndc_minus_one_to_one != 0};
     const float point_size{std::bit_cast<float>(key.state.point_size)};
     switch (stage) {
@@ -185,7 +186,9 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
                     LOG_WARNING(Render_Vulkan, "XFB requested in pipeline key but device lacks VK_EXT_transform_feedback; ignoring XFB decorations");
                 }
             }
-            info.convert_depth_mode = gl_ndc;
+            if (!has_tessellation) {
+                info.convert_depth_mode = gl_ndc;
+            }
         }
         if (key.state.dynamic_vertex_input) {
             for (size_t index = 0; index < Maxwell::NumVertexAttributes; ++index) {
@@ -224,6 +227,9 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
             ASSERT(false);
             return Shader::TessSpacing::Equal;
         }();
+        if (!has_geometry) {
+            info.convert_depth_mode = gl_ndc;
+        }
         break;
     case Shader::Stage::Geometry:
         if (program.output_topology == Shader::OutputTopology::PointList) {
@@ -305,12 +311,8 @@ size_t GetTotalPipelineWorkers() {
         std::max<size_t>(static_cast<size_t>(std::thread::hardware_concurrency()), 2ULL) - 1ULL;
 #ifdef __ANDROID__
     const int configured = AndroidSettings::values.pipeline_worker_count.GetValue();
-    const int clamped = std::clamp(configured, 4, 8);
-    const size_t desired = static_cast<size_t>(clamped);
-    if (desired == 0) {
-        return 1ULL;
-    }
-    return std::min(max_core_threads, desired);
+    const size_t desired = static_cast<size_t>(std::max(configured, 1));
+    return std::min<size_t>(max_core_threads, desired);
 #else
     return max_core_threads;
 #endif

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -241,11 +241,13 @@ void RasterizerVulkan::PrepareDraw(bool is_indexed, Func&& draw_func) {
     if (!pipeline) {
         return;
     }
-    std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
-    // update engine as channel may be different.
-    pipeline->SetEngine(maxwell3d, gpu_memory);
-    if (!pipeline->Configure(is_indexed))
-        return;
+    {
+        std::scoped_lock lock{buffer_cache.mutex, texture_cache.mutex};
+        pipeline->SetEngine(maxwell3d, gpu_memory);
+        if (!pipeline->Configure(is_indexed)) {
+            return;
+        }
+    }
 
     UpdateDynamicStates();
 
@@ -673,7 +675,6 @@ void RasterizerVulkan::FlushRegion(DAddr addr, u64 size, VideoCommon::CacheType
         texture_cache.DownloadMemory(addr, size);
     }
     if ((True(which & VideoCommon::CacheType::BufferCache))) {
-        std::scoped_lock lock{buffer_cache.mutex};
         buffer_cache.DownloadMemory(addr, size);
     }
     if ((True(which & VideoCommon::CacheType::QueryCache))) {
@@ -771,16 +772,22 @@ bool RasterizerVulkan::OnCPUWrite(DAddr addr, u64 size) {
     return false;
 }
 
+static constexpr bool ENABLE_TEXTURE_CACHE_INVALIDATION_SKIP = true;
+static constexpr bool ENABLE_FINE_GRAINED_TRACKER_LOCK = true;
+
 void RasterizerVulkan::OnCacheInvalidation(DAddr addr, u64 size) {
     if (addr == 0 || size == 0) {
         return;
     }
 
-    {
+    if (!ENABLE_TEXTURE_CACHE_INVALIDATION_SKIP ||
+        device_memory.IsRegionTextureCached(addr, size)) {
         std::scoped_lock lock{texture_cache.mutex};
         texture_cache.WriteMemory(addr, size);
     }
-    {
+    if (ENABLE_FINE_GRAINED_TRACKER_LOCK) {
+        buffer_cache.CpuWriteInvalidate(addr, size);
+    } else {
         std::scoped_lock lock{buffer_cache.mutex};
         buffer_cache.WriteMemory(addr, size);
     }

src/video_core/renderer_vulkan/vk_texture_cache.cpp

@@ -129,6 +129,9 @@ constexpr VkBorderColor ConvertBorderColor(const std::array<float, 4>& color) {
     if (info.storage) {
         usage |= VK_IMAGE_USAGE_STORAGE_BIT;
     }
+    if (IsPixelFormatASTC(format)) {
+        usage |= VK_IMAGE_USAGE_COLOR_ATTACHMENT_BIT;
+    }
     return usage;
 }
 
@@ -911,6 +914,10 @@ TextureCacheRuntime::TextureCacheRuntime(const Device& device_, Scheduler& sched
     if (Settings::values.accelerate_astc.GetValue() == Settings::AstcDecodeMode::Gpu) {
         astc_decoder_pass.emplace(device, scheduler, descriptor_pool, staging_buffer_pool,
                                   compute_pass_descriptor_queue, memory_allocator);
+        if (device.IsTiler()) {
+            astc_decoder_fragment_pass.emplace(device, scheduler, descriptor_pool,
+                                               compute_pass_descriptor_queue, render_pass_cache);
+        }
     }
     if (!device.IsKhrImageFormatListSupported()) {
         return;
@@ -2850,6 +2857,13 @@ void TextureCacheRuntime::AccelerateImageUpload(
     u32 z_start, u32 z_count) {
 
     if (IsPixelFormatASTC(image.info.format)) {
+        if (astc_decoder_fragment_pass) {
+            const VideoCore::Surface::PixelFormat decoded_format =
+                WillUseWidenedAstcFormat(device, image.info)
+                    ? VideoCore::Surface::PixelFormat::R32G32B32A32_FLOAT
+                    : VideoCore::Surface::PixelFormat::A8B8G8R8_UNORM;
+            return astc_decoder_fragment_pass->Assemble(image, map, swizzles, decoded_format);
+        }
         return astc_decoder_pass->Assemble(image, map, swizzles);
     }
 

src/video_core/renderer_vulkan/vk_texture_cache.h

@@ -93,7 +93,7 @@ public:
 
     void AccelerateImageUpload(Image&, const StagingBufferRef&,
                                std::span<const VideoCommon::SwizzleParameters>,
-                               u32 z_start, u32 z_count);
+                               u32 z_start = 0, u32 z_count = 0);
 
     void InsertUploadMemoryBarrier() {}
 
@@ -147,6 +147,7 @@ public:
     BlitImageHelper& blit_image_helper;
     RenderPassCache& render_pass_cache;
     std::optional<ASTCDecoderPass> astc_decoder_pass;
+    std::optional<ASTCDecoderFragmentPass> astc_decoder_fragment_pass;
 
     std::optional<BlockLinearUnswizzle3DPass> bl3d_unswizzle_pass;
     const Settings::ResolutionScalingInfo& resolution;

src/video_core/texture_cache/texture_cache.h

@@ -2308,6 +2308,7 @@ void TextureCache<P>::TrackImage(ImageBase& image, ImageId image_id) {
     if (False(image.flags & ImageFlagBits::Sparse)) {
         if (image.cpu_addr < ~(1ULL << 40)) {
             device_memory.UpdatePagesCachedCount(image.cpu_addr, image.guest_size_bytes, 1);
+            device_memory.UpdateTexturePagesCount(image.cpu_addr, image.guest_size_bytes, 1);
         }
         return;
     }
@@ -2320,12 +2321,14 @@ void TextureCache<P>::TrackImage(ImageBase& image, ImageId image_id) {
             const DAddr cpu_addr = map.cpu_addr;
             const std::size_t size = map.size;
             device_memory.UpdatePagesCachedCount(cpu_addr, size, 1);
+            device_memory.UpdateTexturePagesCount(cpu_addr, size, 1);
         }
         return;
     }
     ForEachSparseSegment(image,
                          [this]([[maybe_unused]] GPUVAddr gpu_addr, DAddr cpu_addr, size_t size) {
                              device_memory.UpdatePagesCachedCount(cpu_addr, size, 1);
+                             device_memory.UpdateTexturePagesCount(cpu_addr, size, 1);
                          });
 }
 
@@ -2336,6 +2339,7 @@ void TextureCache<P>::UntrackImage(ImageBase& image, ImageId image_id) {
     if (False(image.flags & ImageFlagBits::Sparse)) {
         if (image.cpu_addr < ~(1ULL << 40)) {
             device_memory.UpdatePagesCachedCount(image.cpu_addr, image.guest_size_bytes, -1);
+            device_memory.UpdateTexturePagesCount(image.cpu_addr, image.guest_size_bytes, -1);
         }
         return;
     }
@@ -2348,6 +2352,7 @@ void TextureCache<P>::UntrackImage(ImageBase& image, ImageId image_id) {
         const DAddr cpu_addr = map.cpu_addr;
         const std::size_t size = map.size;
         device_memory.UpdatePagesCachedCount(cpu_addr, size, -1);
+        device_memory.UpdateTexturePagesCount(cpu_addr, size, -1);
     }
 }
 

src/video_core/texture_cache/texture_cache_base.h

@@ -257,7 +257,7 @@ public:
     /// Prepare an image to be used
     void PrepareImage(ImageId image_id, bool is_modification, bool invalidate);
 
-    std::recursive_mutex mutex;
+    std::mutex mutex;
 
 private:
     /// Iterate over all page indices in a range

src/video_core/vulkan_common/vulkan_device.cpp

@@ -517,16 +517,22 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
                                VK_EXT_COLOR_WRITE_ENABLE_EXTENSION_NAME);
         LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken shader float controls.");
         RemoveExtension(extensions.shader_float_controls, VK_KHR_SHADER_FLOAT_CONTROLS_EXTENSION_NAME);
+        LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken workgroup memory explicit layout.");
+        RemoveExtensionFeature(extensions.workgroup_memory_explicit_layout,
+                               features.workgroup_memory_explicit_layout,
+                               VK_KHR_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_EXTENSION_NAME);
+        LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken conservative rasterization.");
+        RemoveExtension(extensions.conservative_rasterization,
+                        VK_EXT_CONSERVATIVE_RASTERIZATION_EXTENSION_NAME);
+        LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken depth clip control.");
+        RemoveExtensionFeature(extensions.depth_clip_control, features.depth_clip_control,
+                               VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME);
         LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken shader atomic int64.");
         RemoveExtensionFeature(extensions.shader_atomic_int64, features.shader_atomic_int64,
                                VK_KHR_SHADER_ATOMIC_INT64_EXTENSION_NAME);
         features.shader_atomic_int64.shaderBufferInt64Atomics = false;
         features.shader_atomic_int64.shaderSharedInt64Atomics = false;
         features.features.shaderInt64 = false;
-        LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken workgroup memory explicit layout.");
-        RemoveExtensionFeature(extensions.workgroup_memory_explicit_layout,
-                               features.workgroup_memory_explicit_layout,
-                               VK_KHR_WORKGROUP_MEMORY_EXPLICIT_LAYOUT_EXTENSION_NAME);
 
 #if defined(__ANDROID__) && defined(ARCHITECTURE_arm64)
         // BCn patching only safe on Android 9+ (API 28+). Older versions crash on driver load.

src/video_core/vulkan_common/vulkan_wrapper.cpp

@@ -230,6 +230,7 @@ void Load(VkDevice device, DeviceDispatch& dld) noexcept {
     X(vkMapMemory);
     X(vkQueueSubmit);
     X(vkQueueSubmit2);
+    X(vkResetCommandPool);
     X(vkResetFences);
     X(vkResetQueryPool);
     X(vkSetDebugUtilsObjectNameEXT);

src/video_core/vulkan_common/vulkan_wrapper.h

@@ -346,6 +346,7 @@ struct DeviceDispatch : InstanceDispatch {
     PFN_vkMapMemory vkMapMemory{};
     PFN_vkQueueSubmit vkQueueSubmit{};
     PFN_vkQueueSubmit2 vkQueueSubmit2{};
+    PFN_vkResetCommandPool vkResetCommandPool{};
     PFN_vkResetFences vkResetFences{};
     PFN_vkResetQueryPool vkResetQueryPool{};
     PFN_vkSetDebugUtilsObjectNameEXT vkSetDebugUtilsObjectNameEXT{};
@@ -925,6 +926,10 @@ public:
     CommandBuffers Allocate(std::size_t num_buffers,
                             VkCommandBufferLevel level = VK_COMMAND_BUFFER_LEVEL_PRIMARY) const;
 
+    void Reset(VkCommandPoolResetFlags flags = 0) const {
+        Check(dld->vkResetCommandPool(owner, handle, flags));
+    }
+
     /// Set object name.
     void SetObjectNameEXT(const char* name) const;
 };