Android #173

kernel: use simple mutex for object list container

README.md

@@ -1,9 +1,7 @@
 | Pull Request | Commit | Title | Author | Merged? |
 |----|----|----|----|----|
-| [12448](https://github.com/yuzu-emu/yuzu//pull/12448) | [`b1d4804c0`](https://github.com/yuzu-emu/yuzu//pull/12448/files) | renderer_vulkan: demote format assert to error log | [liamwhite](https://github.com/liamwhite/) | Yes |
-| [12449](https://github.com/yuzu-emu/yuzu//pull/12449) | [`6a1ddc502`](https://github.com/yuzu-emu/yuzu//pull/12449/files) | renderer_vulkan: skip SetObjectNameEXT on unsupported driver | [liamwhite](https://github.com/liamwhite/) | Yes |
-| [12466](https://github.com/yuzu-emu/yuzu//pull/12466) | [`5f3720138`](https://github.com/yuzu-emu/yuzu//pull/12466/files) |  core: track separate heap allocation for linux  | [liamwhite](https://github.com/liamwhite/) | Yes |
-| [12467](https://github.com/yuzu-emu/yuzu//pull/12467) | [`cfc6c5f8f`](https://github.com/yuzu-emu/yuzu//pull/12467/files) | Revert " shader_recompiler: use minimal clip distance array " | [liamwhite](https://github.com/liamwhite/) | Yes |
+| [12466](https://github.com/yuzu-emu/yuzu//pull/12466) | [`ddda76f9b`](https://github.com/yuzu-emu/yuzu//pull/12466/files) |  core: track separate heap allocation for linux  | [liamwhite](https://github.com/liamwhite/) | Yes |
+| [12479](https://github.com/yuzu-emu/yuzu//pull/12479) | [`20e040723`](https://github.com/yuzu-emu/yuzu//pull/12479/files) | video_core: Fix buffer_row_length for linear compressed textures | [GPUCode](https://github.com/GPUCode/) | Yes |
 
 
 End of merge log. You can find the original README.md below the break.

src/common/assert.cpp

@@ -3,16 +3,19 @@
 
 #include "common/assert.h"
 #include "common/common_funcs.h"
+#include "common/logging/backend.h"
 
 #include "common/settings.h"
 
 void assert_fail_impl() {
     if (Settings::values.use_debug_asserts) {
+        Common::Log::Stop();
         Crash();
     }
 }
 
 [[noreturn]] void unreachable_impl() {
+    Common::Log::Stop();
     Crash();
     throw std::runtime_error("Unreachable code");
 }

src/common/heap_tracker.cpp

@@ -1,14 +1,17 @@
 // SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include "common/assert.h"
+#include <algorithm>
+#include <vector>
+
 #include "common/heap_tracker.h"
+#include "common/logging/log.h"
 
 namespace Common {
 
 namespace {
 
-constexpr size_t MaxResidentMapCount = 0x8000;
+constexpr s64 MaxResidentMapCount = 0x8000;
 
 } // namespace
 
@@ -27,17 +30,17 @@ void HeapTracker::Map(size_t virtual_offset, size_t host_offset, size_t length,
         // We are mapping part of a separate heap.
         std::scoped_lock lk{m_lock};
 
-        auto* map = new SeparateHeapMap{
+        auto* const map = new SeparateHeapMap{
             .vaddr = virtual_offset,
             .paddr = host_offset,
             .size = length,
-            .map_id = m_next_map_id++,
             .tick = m_tick++,
             .perm = perm,
             .is_resident = false,
         };
 
         // Insert into mappings.
+        m_map_count++;
         m_mappings.insert(*map);
     }
 
@@ -48,11 +51,10 @@ void HeapTracker::Map(size_t virtual_offset, size_t host_offset, size_t length,
 void HeapTracker::Unmap(size_t virtual_offset, size_t size, bool is_separate_heap) {
     // If this is a separate heap...
     if (is_separate_heap) {
-        std::scoped_lock lk{m_rebuild_lock, m_lock};
+        std::scoped_lock lk{m_lock};
 
         const SeparateHeapMap key{
             .vaddr = virtual_offset,
-            .size = size,
         };
 
         // Split at the boundaries of the region we are removing.
@@ -62,20 +64,18 @@ void HeapTracker::Unmap(size_t virtual_offset, size_t size, bool is_separate_hea
         // Erase all mappings in range.
         auto it = m_mappings.find(key);
         while (it != m_mappings.end() && it->vaddr < virtual_offset + size) {
-            // Get pointer to item.
-            SeparateHeapMap* const item = std::addressof(*it);
+            // Get underlying item.
+            auto* const item = std::addressof(*it);
 
+            // If resident, erase from resident map.
             if (item->is_resident) {
-                // Unlink from resident tree.
+                ASSERT(--m_resident_map_count >= 0);
                 m_resident_mappings.erase(m_resident_mappings.iterator_to(*item));
-
-                // Decrease reference count.
-                const auto count_it = m_resident_map_counts.find(item->map_id);
-                this->RemoveReferenceLocked(count_it, 1);
             }
 
-            // Unlink from mapping tree and advance.
+            // Erase from map.
             it = m_mappings.erase(it);
+            ASSERT(--m_map_count >= 0);
 
             // Free the item.
             delete item;
@@ -109,7 +109,7 @@ void HeapTracker::Protect(size_t virtual_offset, size_t size, MemoryPermission p
             };
 
             // Try to get the next mapping corresponding to this address.
-            const auto it = m_mappings.nfind_key(key);
+            const auto it = m_mappings.nfind(key);
 
             if (it == m_mappings.end()) {
                 // There are no separate heap mappings remaining.
@@ -152,106 +152,61 @@ bool HeapTracker::DeferredMapSeparateHeap(u8* fault_address) {
 }
 
 bool HeapTracker::DeferredMapSeparateHeap(size_t virtual_offset) {
-    std::scoped_lock lk{m_lock};
+    bool rebuild_required = false;
 
-    while (this->IsEvictRequiredLocked()) {
-        // Unlock before we rebuild to ensure proper lock ordering.
-        m_lock.unlock();
+    {
+        std::scoped_lock lk{m_lock};
 
-        // Evict four maps.
-        for (size_t i = 0; i < 4; /* ... */) {
-            i += this->EvictSingleSeparateHeapMap();
+        // Check to ensure this was a non-resident separate heap mapping.
+        const auto it = this->GetNearestHeapMapLocked(virtual_offset);
+        if (it == m_mappings.end() || it->is_resident) {
+            return false;
         }
 
-        // Lock again.
-        m_lock.lock();
+        // Update tick before possible rebuild.
+        it->tick = m_tick++;
+
+        // Check if we need to rebuild.
+        if (m_resident_map_count > MaxResidentMapCount) {
+            rebuild_required = true;
+        }
+
+        // Map the area.
+        m_buffer.Map(it->vaddr, it->paddr, it->size, it->perm, false);
+
+        // This map is now resident.
+        it->is_resident = true;
+        m_resident_map_count++;
+        m_resident_mappings.insert(*it);
     }
 
-    // Check to ensure this was a non-resident separate heap mapping.
-    const auto it = this->GetNearestHeapMapLocked(virtual_offset);
-    if (it == m_mappings.end()) {
-        // Not in any separate heap.
-        return false;
-    }
-    if (it->is_resident) {
-        // Already mapped and shouldn't be considered again.
-        return false;
+    if (rebuild_required) {
+        // A rebuild was required, so perform it now.
+        this->RebuildSeparateHeapAddressSpace();
     }
 
-    // Map the area.
-    m_buffer.Map(it->vaddr, it->paddr, it->size, it->perm, false);
-
-    // This map is now resident.
-    this->AddReferenceLocked(it->map_id, 1);
-    it->is_resident = true;
-    it->tick = m_tick++;
-
-    // Insert into resident maps.
-    m_resident_mappings.insert(*it);
-
-    // We succeeded.
     return true;
 }
 
-bool HeapTracker::EvictSingleSeparateHeapMap() {
+void HeapTracker::RebuildSeparateHeapAddressSpace() {
     std::scoped_lock lk{m_rebuild_lock, m_lock};
 
     ASSERT(!m_resident_mappings.empty());
 
-    // Select the item with the lowest tick to evict.
-    auto* const item = std::addressof(*m_resident_mappings.begin());
-    auto it = m_mappings.iterator_to(*item);
+    // Unmap so we have at least 4 maps available.
+    const size_t desired_count = std::min(m_resident_map_count, MaxResidentMapCount - 4);
+    const size_t evict_count = m_resident_map_count - desired_count;
+    auto it = m_resident_mappings.begin();
 
-    // Track the map ID.
-    const size_t map_id = it->map_id;
-
-    // Walk backwards until we find the first entry.
-    while (it != m_mappings.begin()) {
-        // If the previous element does not have the same map ID, stop.
-        const auto prev = std::prev(it);
-        if (prev->map_id != map_id) {
-            break;
-        }
-
-        // Continue.
-        it = prev;
-    }
-
-    // Track the begin and end address.
-    const VAddr begin_vaddr = it->vaddr;
-    VAddr end_vaddr = begin_vaddr;
-
-    // Get the count iterator.
-    const auto count_it = m_resident_map_counts.find(map_id);
-
-    // Declare whether we have erased an underlying mapping.
-    bool was_erased = false;
-
-    // Unmark and merge everything in range.
-    while (it != m_mappings.end() && it->map_id == map_id) {
-        if (it->is_resident) {
-            // Remove from resident tree.
-            m_resident_mappings.erase(m_resident_mappings.iterator_to(*it));
-            it->is_resident = false;
-
-            // Remove reference count.
-            was_erased |= this->RemoveReferenceLocked(count_it, 1);
-        }
-
-        // Update the end address.
-        end_vaddr = it->vaddr + it->size;
+    for (size_t i = 0; i < evict_count && it != m_resident_mappings.end(); i++) {
+        // Unmark and unmap.
+        it->is_resident = false;
+        m_buffer.Unmap(it->vaddr, it->size, false);
 
         // Advance.
-        it = this->MergeHeapMapForEvictLocked(it);
+        ASSERT(--m_resident_map_count >= 0);
+        it = m_resident_mappings.erase(it);
     }
-
-    // Finally, unmap.
-    ASSERT(end_vaddr >= begin_vaddr);
-    m_buffer.Unmap(begin_vaddr, end_vaddr - begin_vaddr, false);
-
-    // Return whether we actually removed a mapping.
-    // This will be true if there were no holes, which is likely.
-    return was_erased;
 }
 
 void HeapTracker::SplitHeapMap(VAddr offset, size_t size) {
@@ -268,11 +223,9 @@ void HeapTracker::SplitHeapMapLocked(VAddr offset) {
         return;
     }
 
-    // Get the underlying item as the left.
+    // Cache the original values.
     auto* const left = std::addressof(*it);
-
-    // Cache the original size values.
-    const size_t size = left->size;
+    const size_t orig_size = left->size;
 
     // Adjust the left map.
     const size_t left_size = offset - left->vaddr;
@@ -282,74 +235,21 @@ void HeapTracker::SplitHeapMapLocked(VAddr offset) {
     auto* const right = new SeparateHeapMap{
         .vaddr = left->vaddr + left_size,
         .paddr = left->paddr + left_size,
-        .size = size - left_size,
-        .map_id = left->map_id,
+        .size = orig_size - left_size,
         .tick = left->tick,
         .perm = left->perm,
         .is_resident = left->is_resident,
     };
 
     // Insert the new right map.
+    m_map_count++;
     m_mappings.insert(*right);
 
-    // If the original map was not resident, we are done.
-    if (!left->is_resident) {
-        return;
+    // If resident, also insert into resident map.
+    if (right->is_resident) {
+        m_resident_mappings.insert(*right);
+        m_resident_map_count++;
     }
-
-    // Update reference count.
-    this->AddReferenceLocked(left->map_id, 1);
-
-    // Insert right into resident map.
-    m_resident_mappings.insert(*right);
-}
-
-HeapTracker::AddrTree::iterator HeapTracker::MergeHeapMapForEvictLocked(AddrTree::iterator it) {
-    if (it == m_mappings.end()) {
-        // Not contained.
-        return it;
-    }
-
-    if (it == m_mappings.begin()) {
-        // Nothing to merge with.
-        return std::next(it);
-    }
-
-    // Get the left and right items.
-    auto* const right = std::addressof(*it);
-    auto* const left = std::addressof(*std::prev(it));
-
-    if (left->vaddr + left->size != right->vaddr) {
-        // Virtual range not contiguous, cannot merge.
-        return std::next(it);
-    }
-
-    if (left->paddr + left->size != right->paddr) {
-        // Physical range not contiguous, cannot merge.
-        return std::next(it);
-    }
-
-    if (left->perm != right->perm) {
-        // Permissions mismatch, cannot merge.
-        return std::next(it);
-    }
-
-    if (left->map_id != right->map_id) {
-        // Map ID mismatch, cannot merge.
-        return std::next(it);
-    }
-
-    // Merge size to the left.
-    left->size += right->size;
-
-    // Erase the right element.
-    const auto next_it = m_mappings.erase(it);
-
-    // Free the right element.
-    delete right;
-
-    // Return the iterator to the next position.
-    return next_it;
 }
 
 HeapTracker::AddrTree::iterator HeapTracker::GetNearestHeapMapLocked(VAddr offset) {
@@ -360,26 +260,4 @@ HeapTracker::AddrTree::iterator HeapTracker::GetNearestHeapMapLocked(VAddr offse
     return m_mappings.find(key);
 }
 
-void HeapTracker::AddReferenceLocked(size_t map_id, size_t inc) {
-    m_resident_map_counts[map_id]++;
-}
-
-bool HeapTracker::RemoveReferenceLocked(MapCountTree::iterator it, size_t dec) {
-    ASSERT(it != m_resident_map_counts.end());
-
-    const auto new_value = it->second -= dec;
-    ASSERT(new_value >= 0);
-
-    if (new_value <= 0) {
-        m_resident_map_counts.erase(it);
-        return true;
-    }
-
-    return false;
-}
-
-bool HeapTracker::IsEvictRequiredLocked() {
-    return m_resident_map_counts.size() > MaxResidentMapCount;
-}
-
 } // namespace Common

src/common/heap_tracker.h

@@ -3,8 +3,9 @@
 
 #pragma once
 
-#include <map>
+#include <atomic>
 #include <mutex>
+#include <set>
 #include <shared_mutex>
 
 #include "common/host_memory.h"
@@ -18,7 +19,6 @@ struct SeparateHeapMap {
     VAddr vaddr{};
     PAddr paddr{};
     size_t size{};
-    size_t map_id{};
     size_t tick{};
     MemoryPermission perm{};
     bool is_resident{};
@@ -64,14 +64,6 @@ public:
     bool DeferredMapSeparateHeap(u8* fault_address);
     bool DeferredMapSeparateHeap(size_t virtual_offset);
 
-private:
-    Common::HostMemory& m_buffer;
-
-    std::shared_mutex m_rebuild_lock{};
-    std::mutex m_lock{};
-    size_t m_next_map_id{};
-    size_t m_tick{};
-
 private:
     using AddrTreeTraits =
         Common::IntrusiveRedBlackTreeMemberTraitsDeferredAssert<&SeparateHeapMap::addr_node>;
@@ -80,9 +72,7 @@ private:
     using TickTreeTraits =
         Common::IntrusiveRedBlackTreeMemberTraitsDeferredAssert<&SeparateHeapMap::tick_node>;
     using TickTree = TickTreeTraits::TreeType<SeparateHeapMapTickComparator>;
-    using MapCountTree = std::map<size_t, s64>;
 
-    MapCountTree m_resident_map_counts{};
     AddrTree m_mappings{};
     TickTree m_resident_mappings{};
 
@@ -90,14 +80,18 @@ private:
     void SplitHeapMap(VAddr offset, size_t size);
     void SplitHeapMapLocked(VAddr offset);
 
-    AddrTree::iterator MergeHeapMapForEvictLocked(AddrTree::iterator cur);
     AddrTree::iterator GetNearestHeapMapLocked(VAddr offset);
 
-    bool EvictSingleSeparateHeapMap();
+    void RebuildSeparateHeapAddressSpace();
 
-    void AddReferenceLocked(size_t map_id, size_t inc);
-    bool RemoveReferenceLocked(MapCountTree::iterator map_id, size_t dec);
-    bool IsEvictRequiredLocked();
+private:
+    Common::HostMemory& m_buffer;
+
+    std::shared_mutex m_rebuild_lock{};
+    std::mutex m_lock{};
+    s64 m_map_count{};
+    s64 m_resident_map_count{};
+    size_t m_tick{};
 };
 
 } // namespace Common

src/common/logging/backend.cpp

@@ -208,6 +208,10 @@ public:
         instance->StartBackendThread();
     }
 
+    static void Stop() {
+        instance->StopBackendThread();
+    }
+
     Impl(const Impl&) = delete;
     Impl& operator=(const Impl&) = delete;
 
@@ -259,6 +263,15 @@ private:
         });
     }
 
+    void StopBackendThread() {
+        backend_thread.request_stop();
+        if (backend_thread.joinable()) {
+            backend_thread.join();
+        }
+
+        ForEachBackend([](Backend& backend) { backend.Flush(); });
+    }
+
     Entry CreateEntry(Class log_class, Level log_level, const char* filename, unsigned int line_nr,
                       const char* function, std::string&& message) const {
         using std::chrono::duration_cast;
@@ -313,6 +326,10 @@ void Start() {
     Impl::Start();
 }
 
+void Stop() {
+    Impl::Stop();
+}
+
 void DisableLoggingInTests() {
     initialization_in_progress_suppress_logging = true;
 }

src/common/logging/backend.h

@@ -14,6 +14,9 @@ void Initialize();
 
 void Start();
 
+/// Explicitly stops the logger thread and flushes the buffers
+void Stop();
+
 void DisableLoggingInTests();
 
 /**

src/common/ring_buffer.h

@@ -103,7 +103,7 @@ private:
     // Having them on the same cache-line would result in false-sharing between them.
     // TODO: Remove this ifdef whenever clang and GCC support
     //       std::hardware_destructive_interference_size.
-#if defined(_MSC_VER) && _MSC_VER >= 1911
+#ifdef __cpp_lib_hardware_interference_size
     alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_read_index{0};
     alignas(std::hardware_destructive_interference_size) std::atomic_size_t m_write_index{0};
 #else

src/core/hle/kernel/k_auto_object_container.cpp

@@ -8,19 +8,22 @@
 namespace Kernel {
 
 void KAutoObjectWithListContainer::Register(KAutoObjectWithList* obj) {
-    KScopedLightLock lk(m_lock);
+    // KScopedInterruptDisable di;
+    KScopedSpinLock lk(m_lock);
 
     m_object_list.insert_unique(*obj);
 }
 
 void KAutoObjectWithListContainer::Unregister(KAutoObjectWithList* obj) {
-    KScopedLightLock lk(m_lock);
+    // KScopedInterruptDisable di;
+    KScopedSpinLock lk(m_lock);
 
     m_object_list.erase(*obj);
 }
 
 size_t KAutoObjectWithListContainer::GetOwnedCount(KProcess* owner) {
-    KScopedLightLock lk(m_lock);
+    // KScopedInterruptDisable di;
+    KScopedSpinLock lk(m_lock);
 
     return std::count_if(m_object_list.begin(), m_object_list.end(),
                          [&](const auto& obj) { return obj.GetOwner() == owner; });

src/core/hle/kernel/k_auto_object_container.h

@@ -7,7 +7,7 @@
 
 #include "common/common_funcs.h"
 #include "core/hle/kernel/k_auto_object.h"
-#include "core/hle/kernel/k_light_lock.h"
+#include "core/hle/kernel/k_spin_lock.h"
 
 namespace Kernel {
 
@@ -21,32 +21,7 @@ public:
 
     using ListType = boost::intrusive::rbtree<KAutoObjectWithList>;
 
-    class ListAccessor : public KScopedLightLock {
-    public:
-        explicit ListAccessor(KAutoObjectWithListContainer* container)
-            : KScopedLightLock(container->m_lock), m_list(container->m_object_list) {}
-        explicit ListAccessor(KAutoObjectWithListContainer& container)
-            : KScopedLightLock(container.m_lock), m_list(container.m_object_list) {}
-
-        typename ListType::iterator begin() const {
-            return m_list.begin();
-        }
-
-        typename ListType::iterator end() const {
-            return m_list.end();
-        }
-
-        typename ListType::iterator find(typename ListType::const_reference ref) const {
-            return m_list.find(ref);
-        }
-
-    private:
-        ListType& m_list;
-    };
-
-    friend class ListAccessor;
-
-    KAutoObjectWithListContainer(KernelCore& kernel) : m_lock(kernel), m_object_list() {}
+    KAutoObjectWithListContainer(KernelCore& kernel) : m_lock(), m_object_list() {}
 
     void Initialize() {}
     void Finalize() {}
@@ -56,7 +31,7 @@ public:
     size_t GetOwnedCount(KProcess* owner);
 
 private:
-    KLightLock m_lock;
+    KSpinLock m_lock;
     ListType m_object_list;
 };
 

src/core/hle/kernel/k_handle_table.h

@@ -30,7 +30,7 @@ public:
 public:
     explicit KHandleTable(KernelCore& kernel) : m_kernel(kernel) {}
 
-    Result Initialize(KProcess* owner, s32 size) {
+    Result Initialize(s32 size) {
         // Check that the table size is valid.
         R_UNLESS(size <= static_cast<s32>(MaxTableSize), ResultOutOfMemory);
 
@@ -44,7 +44,6 @@ public:
         m_next_linear_id = MinLinearId;
         m_count = 0;
         m_free_head_index = -1;
-        m_owner = owner;
 
         // Free all entries.
         for (s32 i = 0; i < static_cast<s32>(m_table_size); ++i) {
@@ -91,8 +90,7 @@ public:
         // Handle pseudo-handles.
         if constexpr (std::derived_from<KProcess, T>) {
             if (handle == Svc::PseudoHandle::CurrentProcess) {
-                // TODO: this should be the current process
-                auto* const cur_process = m_owner;
+                auto* const cur_process = GetCurrentProcessPointer(m_kernel);
                 ASSERT(cur_process != nullptr);
                 return cur_process;
             }
@@ -302,7 +300,6 @@ private:
 
 private:
     KernelCore& m_kernel;
-    KProcess* m_owner{};
     std::array<EntryInfo, MaxTableSize> m_entry_infos{};
     std::array<KAutoObject*, MaxTableSize> m_objects{};
     mutable KSpinLock m_lock;

src/core/hle/kernel/k_process.h

@@ -552,7 +552,7 @@ private:
 
     Result InitializeHandleTable(s32 size) {
         // Try to initialize the handle table.
-        R_TRY(m_handle_table.Initialize(this, size));
+        R_TRY(m_handle_table.Initialize(size));
 
         // We succeeded, so note that we did.
         m_is_handle_table_initialized = true;

src/core/hle/kernel/k_server_session.cpp

@@ -1147,8 +1147,7 @@ Result KServerSession::ReceiveRequest(uintptr_t server_message, uintptr_t server
         *out_context =
             std::make_shared<Service::HLERequestContext>(m_kernel, memory, this, client_thread);
         (*out_context)->SetSessionRequestManager(manager);
-        (*out_context)
-            ->PopulateFromIncomingCommandBuffer(*client_thread->GetOwnerProcess(), cmd_buf);
+        (*out_context)->PopulateFromIncomingCommandBuffer(cmd_buf);
         // We succeeded.
         R_SUCCEED();
     } else {

src/core/hle/kernel/k_transfer_memory.h

@@ -5,6 +5,7 @@
 
 #include <optional>
 
+#include "core/hle/kernel/k_light_lock.h"
 #include "core/hle/kernel/k_page_group.h"
 #include "core/hle/kernel/slab_helpers.h"
 #include "core/hle/kernel/svc_types.h"

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

@@ -1513,8 +1513,7 @@ void ILibraryAppletCreator::CreateTransferMemoryStorage(HLERequestContext& ctx)
         return;
     }
 
-    auto transfer_mem =
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(handle);
+    auto transfer_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(handle);
 
     if (transfer_mem.IsNull()) {
         LOG_ERROR(Service_AM, "transfer_mem is a nullptr for handle={:08X}", handle);
@@ -1524,8 +1523,7 @@ void ILibraryAppletCreator::CreateTransferMemoryStorage(HLERequestContext& ctx)
     }
 
     std::vector<u8> memory(transfer_mem->GetSize());
-    system.ApplicationMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(),
-                                         memory.size());
+    ctx.GetMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(), memory.size());
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(ResultSuccess);
@@ -1547,8 +1545,7 @@ void ILibraryAppletCreator::CreateHandleStorage(HLERequestContext& ctx) {
         return;
     }
 
-    auto transfer_mem =
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(handle);
+    auto transfer_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(handle);
 
     if (transfer_mem.IsNull()) {
         LOG_ERROR(Service_AM, "transfer_mem is a nullptr for handle={:08X}", handle);
@@ -1558,8 +1555,7 @@ void ILibraryAppletCreator::CreateHandleStorage(HLERequestContext& ctx) {
     }
 
     std::vector<u8> memory(transfer_mem->GetSize());
-    system.ApplicationMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(),
-                                         memory.size());
+    ctx.GetMemory().ReadBlock(transfer_mem->GetSourceAddress(), memory.data(), memory.size());
 
     IPC::ResponseBuilder rb{ctx, 2, 0, 1};
     rb.Push(ResultSuccess);

src/core/hle/service/audio/audren_u.cpp

@@ -454,10 +454,8 @@ void AudRenU::OpenAudioRenderer(HLERequestContext& ctx) {
         return;
     }
 
-    const auto& handle_table{system.ApplicationProcess()->GetHandleTable()};
-    auto process{handle_table.GetObject<Kernel::KProcess>(process_handle)};
-    auto transfer_memory{
-        process->GetHandleTable().GetObject<Kernel::KTransferMemory>(transfer_memory_handle)};
+    auto process{ctx.GetObjectFromHandle<Kernel::KProcess>(process_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     const auto session_id{impl->GetSessionId()};
     if (session_id == -1) {

src/core/hle/service/audio/hwopus.cpp

@@ -278,9 +278,7 @@ void HwOpus::OpenHardwareOpusDecoder(HLERequestContext& ctx) {
     auto params = rp.PopRaw<OpusParameters>();
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio, "sample_rate {} channel_count {} transfer_memory_size 0x{:X}",
               params.sample_rate, params.channel_count, transfer_memory_size);
@@ -323,9 +321,7 @@ void HwOpus::OpenHardwareOpusDecoderForMultiStream(HLERequestContext& ctx) {
 
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio,
               "sample_rate {} channel_count {} total_stream_count {} stereo_stream_count {} "
@@ -374,9 +370,7 @@ void HwOpus::OpenHardwareOpusDecoderEx(HLERequestContext& ctx) {
     auto params = rp.PopRaw<OpusParametersEx>();
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio, "sample_rate {} channel_count {} transfer_memory_size 0x{:X}",
               params.sample_rate, params.channel_count, transfer_memory_size);
@@ -414,9 +408,7 @@ void HwOpus::OpenHardwareOpusDecoderForMultiStreamEx(HLERequestContext& ctx) {
 
     auto transfer_memory_size{rp.Pop<u32>()};
     auto transfer_memory_handle{ctx.GetCopyHandle(0)};
-    auto transfer_memory{
-        system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-            transfer_memory_handle)};
+    auto transfer_memory{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(transfer_memory_handle)};
 
     LOG_DEBUG(Service_Audio,
               "sample_rate {} channel_count {} total_stream_count {} stereo_stream_count {} "

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

@@ -1850,8 +1850,7 @@ void IHidServer::InitializeSevenSixAxisSensor(HLERequestContext& ctx) {
     ASSERT_MSG(t_mem_1_size == 0x1000, "t_mem_1_size is not 0x1000 bytes");
     ASSERT_MSG(t_mem_2_size == 0x7F000, "t_mem_2_size is not 0x7F000 bytes");
 
-    auto t_mem_1 = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_1_handle);
+    auto t_mem_1 = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_1_handle);
 
     if (t_mem_1.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem_1 is a nullptr for handle=0x{:08X}", t_mem_1_handle);
@@ -1860,8 +1859,7 @@ void IHidServer::InitializeSevenSixAxisSensor(HLERequestContext& ctx) {
         return;
     }
 
-    auto t_mem_2 = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_2_handle);
+    auto t_mem_2 = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_2_handle);
 
     if (t_mem_2.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem_2 is a nullptr for handle=0x{:08X}", t_mem_2_handle);
@@ -2142,8 +2140,7 @@ void IHidServer::WritePalmaWaveEntry(HLERequestContext& ctx) {
 
     ASSERT_MSG(t_mem_size == 0x3000, "t_mem_size is not 0x3000 bytes");
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     if (t_mem.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem is a nullptr for handle=0x{:08X}", t_mem_handle);

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

@@ -448,8 +448,7 @@ void HidBus::EnableJoyPollingReceiveMode(HLERequestContext& ctx) {
 
     ASSERT_MSG(t_mem_size == 0x1000, "t_mem_size is not 0x1000 bytes");
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     if (t_mem.IsNull()) {
         LOG_ERROR(Service_HID, "t_mem is a nullptr for handle=0x{:08X}", t_mem_handle);

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

@@ -197,8 +197,7 @@ void IRS::RunImageTransferProcessor(HLERequestContext& ctx) {
     const auto parameters{rp.PopRaw<Parameters>()};
     const auto t_mem_handle{ctx.GetCopyHandle(0)};
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     if (t_mem.IsNull()) {
         LOG_ERROR(Service_IRS, "t_mem is a nullptr for handle=0x{:08X}", t_mem_handle);
@@ -444,8 +443,7 @@ void IRS::RunImageTransferExProcessor(HLERequestContext& ctx) {
     const auto parameters{rp.PopRaw<Parameters>()};
     const auto t_mem_handle{ctx.GetCopyHandle(0)};
 
-    auto t_mem = system.ApplicationProcess()->GetHandleTable().GetObject<Kernel::KTransferMemory>(
-        t_mem_handle);
+    auto t_mem = ctx.GetObjectFromHandle<Kernel::KTransferMemory>(t_mem_handle);
 
     LOG_INFO(Service_IRS,
              "called, npad_type={}, npad_id={}, transfer_memory_size={}, "

src/core/hle/service/hle_ipc.cpp

@@ -146,10 +146,7 @@ HLERequestContext::HLERequestContext(Kernel::KernelCore& kernel_, Core::Memory::
 
 HLERequestContext::~HLERequestContext() = default;
 
-void HLERequestContext::ParseCommandBuffer(Kernel::KProcess& process, u32_le* src_cmdbuf,
-                                           bool incoming) {
-    client_handle_table = &process.GetHandleTable();
-
+void HLERequestContext::ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming) {
     IPC::RequestParser rp(src_cmdbuf);
     command_header = rp.PopRaw<IPC::CommandHeader>();
 
@@ -162,7 +159,7 @@ void HLERequestContext::ParseCommandBuffer(Kernel::KProcess& process, u32_le* sr
     if (command_header->enable_handle_descriptor) {
         handle_descriptor_header = rp.PopRaw<IPC::HandleDescriptorHeader>();
         if (handle_descriptor_header->send_current_pid) {
-            pid = process.GetProcessId();
+            pid = thread->GetOwnerProcess()->GetProcessId();
             rp.Skip(2, false);
         }
         if (incoming) {
@@ -270,9 +267,10 @@ void HLERequestContext::ParseCommandBuffer(Kernel::KProcess& process, u32_le* sr
     rp.Skip(1, false); // The command is actually an u64, but we don't use the high part.
 }
 
-Result HLERequestContext::PopulateFromIncomingCommandBuffer(Kernel::KProcess& process,
-                                                            u32_le* src_cmdbuf) {
-    ParseCommandBuffer(process, src_cmdbuf, true);
+Result HLERequestContext::PopulateFromIncomingCommandBuffer(u32_le* src_cmdbuf) {
+    client_handle_table = &thread->GetOwnerProcess()->GetHandleTable();
+
+    ParseCommandBuffer(src_cmdbuf, true);
 
     if (command_header->IsCloseCommand()) {
         // Close does not populate the rest of the IPC header
@@ -284,9 +282,9 @@ Result HLERequestContext::PopulateFromIncomingCommandBuffer(Kernel::KProcess& pr
     return ResultSuccess;
 }
 
-Result HLERequestContext::WriteToOutgoingCommandBuffer(Kernel::KThread& requesting_thread) {
+Result HLERequestContext::WriteToOutgoingCommandBuffer() {
     auto current_offset = handles_offset;
-    auto& owner_process = *requesting_thread.GetOwnerProcess();
+    auto& owner_process = *thread->GetOwnerProcess();
     auto& handle_table = owner_process.GetHandleTable();
 
     for (auto& object : outgoing_copy_objects) {
@@ -319,7 +317,7 @@ Result HLERequestContext::WriteToOutgoingCommandBuffer(Kernel::KThread& requesti
     }
 
     // Copy the translated command buffer back into the thread's command buffer area.
-    memory.WriteBlock(requesting_thread.GetTlsAddress(), cmd_buf.data(), write_size * sizeof(u32));
+    memory.WriteBlock(thread->GetTlsAddress(), cmd_buf.data(), write_size * sizeof(u32));
 
     return ResultSuccess;
 }

src/core/hle/service/hle_ipc.h

@@ -17,6 +17,7 @@
 #include "common/concepts.h"
 #include "common/swap.h"
 #include "core/hle/ipc.h"
+#include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/svc_common.h"
 
 union Result;
@@ -196,10 +197,10 @@ public:
     }
 
     /// Populates this context with data from the requesting process/thread.
-    Result PopulateFromIncomingCommandBuffer(Kernel::KProcess& process, u32_le* src_cmdbuf);
+    Result PopulateFromIncomingCommandBuffer(u32_le* src_cmdbuf);
 
     /// Writes data from this context back to the requesting process/thread.
-    Result WriteToOutgoingCommandBuffer(Kernel::KThread& requesting_thread);
+    Result WriteToOutgoingCommandBuffer();
 
     [[nodiscard]] u32_le GetHipcCommand() const {
         return command;
@@ -359,8 +360,17 @@ public:
         return *thread;
     }
 
-    Kernel::KHandleTable& GetClientHandleTable() {
-        return *client_handle_table;
+    [[nodiscard]] Core::Memory::Memory& GetMemory() const {
+        return memory;
+    }
+
+    template <typename T>
+    Kernel::KScopedAutoObject<T> GetObjectFromHandle(u32 handle) {
+        auto obj = client_handle_table->GetObjectForIpc(handle, thread);
+        if (obj.IsNotNull()) {
+            return obj->DynamicCast<T*>();
+        }
+        return nullptr;
     }
 
     [[nodiscard]] std::shared_ptr<SessionRequestManager> GetManager() const {
@@ -378,7 +388,7 @@ public:
 private:
     friend class IPC::ResponseBuilder;
 
-    void ParseCommandBuffer(Kernel::KProcess& process, u32_le* src_cmdbuf, bool incoming);
+    void ParseCommandBuffer(u32_le* src_cmdbuf, bool incoming);
 
     std::array<u32, IPC::COMMAND_BUFFER_LENGTH> cmd_buf;
     Kernel::KServerSession* server_session{};

src/core/hle/service/jit/jit.cpp

@@ -26,7 +26,7 @@ public:
     explicit IJitEnvironment(Core::System& system_, Kernel::KProcess& process_, CodeRange user_rx,
                              CodeRange user_ro)
         : ServiceFramework{system_, "IJitEnvironment"}, process{&process_},
-          context{system_.ApplicationMemory()} {
+          context{process->GetMemory()} {
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &IJitEnvironment::GenerateCode, "GenerateCode"},
@@ -188,7 +188,7 @@ public:
             return;
         }
 
-        auto tmem{process->GetHandleTable().GetObject<Kernel::KTransferMemory>(tmem_handle)};
+        auto tmem{ctx.GetObjectFromHandle<Kernel::KTransferMemory>(tmem_handle)};
         if (tmem.IsNull()) {
             LOG_ERROR(Service_JIT, "attempted to load plugin with invalid transfer memory handle");
             IPC::ResponseBuilder rb{ctx, 2};
@@ -356,11 +356,7 @@ public:
             return;
         }
 
-        // Fetch using the handle table for the application process here,
-        // since we are not multiprocess yet.
-        const auto& handle_table{system.ApplicationProcess()->GetHandleTable()};
-
-        auto process{handle_table.GetObject<Kernel::KProcess>(process_handle)};
+        auto process{ctx.GetObjectFromHandle<Kernel::KProcess>(process_handle)};
         if (process.IsNull()) {
             LOG_ERROR(Service_JIT, "process is null for handle=0x{:08X}", process_handle);
             IPC::ResponseBuilder rb{ctx, 2};
@@ -368,7 +364,7 @@ public:
             return;
         }
 
-        auto rx_mem{handle_table.GetObject<Kernel::KCodeMemory>(rx_mem_handle)};
+        auto rx_mem{ctx.GetObjectFromHandle<Kernel::KCodeMemory>(rx_mem_handle)};
         if (rx_mem.IsNull()) {
             LOG_ERROR(Service_JIT, "rx_mem is null for handle=0x{:08X}", rx_mem_handle);
             IPC::ResponseBuilder rb{ctx, 2};
@@ -376,7 +372,7 @@ public:
             return;
         }
 
-        auto ro_mem{handle_table.GetObject<Kernel::KCodeMemory>(ro_mem_handle)};
+        auto ro_mem{ctx.GetObjectFromHandle<Kernel::KCodeMemory>(ro_mem_handle)};
         if (ro_mem.IsNull()) {
             LOG_ERROR(Service_JIT, "ro_mem is null for handle=0x{:08X}", ro_mem_handle);
             IPC::ResponseBuilder rb{ctx, 2};

src/core/hle/service/ro/ro.cpp

@@ -651,10 +651,9 @@ private:
     void RegisterProcessHandle(HLERequestContext& ctx) {
         LOG_DEBUG(Service_LDR, "(called)");
 
-        auto process_h = ctx.GetClientHandleTable().GetObject(ctx.GetCopyHandle(0));
+        auto process = ctx.GetObjectFromHandle<Kernel::KProcess>(ctx.GetCopyHandle(0));
         auto client_pid = ctx.GetPID();
-        auto result = interface.RegisterProcessHandle(client_pid,
-                                                      process_h->DynamicCast<Kernel::KProcess*>());
+        auto result = interface.RegisterProcessHandle(client_pid, process.GetPointerUnsafe());
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);
@@ -671,12 +670,11 @@ private:
 
         IPC::RequestParser rp{ctx};
         auto params = rp.PopRaw<InputParameters>();
-        auto process_h = ctx.GetClientHandleTable().GetObject(ctx.GetCopyHandle(0));
+        auto process = ctx.GetObjectFromHandle<Kernel::KProcess>(ctx.GetCopyHandle(0));
 
         auto client_pid = ctx.GetPID();
-        auto result =
-            interface.RegisterProcessModuleInfo(client_pid, params.nrr_address, params.nrr_size,
-                                                process_h->DynamicCast<Kernel::KProcess*>());
+        auto result = interface.RegisterProcessModuleInfo(
+            client_pid, params.nrr_address, params.nrr_size, process.GetPointerUnsafe());
 
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(result);

src/core/hle/service/service.cpp

@@ -203,7 +203,7 @@ Result ServiceFrameworkBase::HandleSyncRequest(Kernel::KServerSession& session,
     // If emulation was shutdown, we are closing service threads, do not write the response back to
     // memory that may be shutting down as well.
     if (system.IsPoweredOn()) {
-        ctx.WriteToOutgoingCommandBuffer(ctx.GetThread());
+        ctx.WriteToOutgoingCommandBuffer();
     }
 
     return result;

src/core/memory.cpp

@@ -730,17 +730,6 @@ struct Memory::Impl {
             GetInteger(vaddr), []() {}, []() {});
     }
 
-    void FixPageProtection(u64 vaddr) {
-        vaddr = Common::AlignDown(vaddr, YUZU_PAGESIZE);
-
-        if (!AddressSpaceContains(*current_page_table, vaddr, 1)) [[unlikely]] {
-            return;
-        }
-
-        ProtectRegion(*current_page_table, vaddr, YUZU_PAGESIZE,
-                      Common::MemoryPermission::ReadWrite);
-    }
-
     /**
      * Reads a particular data type out of memory at the given virtual address.
      *
@@ -1094,18 +1083,13 @@ bool Memory::InvalidateNCE(Common::ProcessAddress vaddr, size_t size) {
             rasterizer = true;
         });
 
-    const bool mapping_exists = mapped && ptr != nullptr;
-
 #ifdef __linux__
-    if (mapping_exists && !rasterizer) {
-        if (!impl->buffer->DeferredMapSeparateHeap(GetInteger(vaddr))) {
-            // GPU may have raced reprotecting this page, try to fix it.
-            impl->FixPageProtection(GetInteger(vaddr));
-        }
+    if (!rasterizer && mapped) {
+        impl->buffer->DeferredMapSeparateHeap(GetInteger(vaddr));
     }
 #endif
 
-    return mapping_exists;
+    return mapped && ptr != nullptr;
 }
 
 bool Memory::InvalidateSeparateHeap(void* fault_address) {

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -74,6 +74,11 @@ std::optional<OutAttr> OutputAttrPointer(EmitContext& ctx, IR::Attribute attr) {
     case IR::Attribute::ClipDistance7: {
         const u32 base{static_cast<u32>(IR::Attribute::ClipDistance0)};
         const u32 index{static_cast<u32>(attr) - base};
+        if (index >= ctx.profile.max_user_clip_distances) {
+            LOG_WARNING(Shader, "Ignoring clip distance store {} >= {} supported", index,
+                        ctx.profile.max_user_clip_distances);
+            return std::nullopt;
+        }
         const Id clip_num{ctx.Const(index)};
         return OutputAccessChain(ctx, ctx.output_f32, ctx.clip_distances, clip_num);
     }

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -1532,7 +1532,8 @@ void EmitContext::DefineOutputs(const IR::Program& program) {
         if (stage == Stage::Fragment) {
             throw NotImplementedException("Storing ClipDistance in fragment stage");
         }
-        const Id type{TypeArray(F32[1], Const(8U))};
+        const Id type{TypeArray(
+            F32[1], Const(std::min(info.used_clip_distances, profile.max_user_clip_distances)))};
         clip_distances = DefineOutput(*this, type, invocations, spv::BuiltIn::ClipDistance);
     }
     if (info.stores[IR::Attribute::Layer] &&

src/shader_recompiler/ir_opt/collect_shader_info_pass.cpp

@@ -913,7 +913,11 @@ void GatherInfoFromHeader(Environment& env, Info& info) {
         }
         for (size_t index = 0; index < 8; ++index) {
             const u16 mask{header.vtg.omap_systemc.clip_distances};
-            info.stores.Set(IR::Attribute::ClipDistance0 + index, ((mask >> index) & 1) != 0);
+            const bool used{((mask >> index) & 1) != 0};
+            info.stores.Set(IR::Attribute::ClipDistance0 + index, used);
+            if (used) {
+                info.used_clip_distances = static_cast<u32>(index) + 1;
+            }
         }
         info.stores.Set(IR::Attribute::PrimitiveId,
                         header.vtg.omap_systemb.primitive_array_id != 0);

src/shader_recompiler/profile.h

@@ -87,6 +87,8 @@ struct Profile {
     bool has_broken_robust{};
 
     u64 min_ssbo_alignment{};
+
+    u32 max_user_clip_distances{};
 };
 
 } // namespace Shader

src/shader_recompiler/shader_info.h

@@ -324,6 +324,8 @@ struct Info {
     bool requires_layer_emulation{};
     IR::Attribute emulated_layer{};
 
+    u32 used_clip_distances{};
+
     boost::container::static_vector<ConstantBufferDescriptor, MAX_CBUFS>
         constant_buffer_descriptors;
     boost::container::static_vector<StorageBufferDescriptor, MAX_SSBOS> storage_buffers_descriptors;

src/video_core/macro/macro_hle.cpp

@@ -327,12 +327,13 @@ public:
     explicit HLE_DrawIndirectByteCount(Maxwell3D& maxwell3d_) : HLEMacroImpl(maxwell3d_) {}
 
     void Execute(const std::vector<u32>& parameters, [[maybe_unused]] u32 method) override {
+        const bool force = maxwell3d.Rasterizer().HasDrawTransformFeedback();
+
         auto topology = static_cast<Maxwell3D::Regs::PrimitiveTopology>(parameters[0] & 0xFFFFU);
-        if (!maxwell3d.AnyParametersDirty() || !IsTopologySafe(topology)) {
+        if (!force && (!maxwell3d.AnyParametersDirty() || !IsTopologySafe(topology))) {
             Fallback(parameters);
             return;
         }
-
         auto& params = maxwell3d.draw_manager->GetIndirectParams();
         params.is_byte_count = true;
         params.is_indexed = false;
@@ -503,6 +504,8 @@ public:
         maxwell3d.CallMethod(static_cast<size_t>(MAXWELL3D_REG_INDEX(launch_dma)), 0x1011, true);
         maxwell3d.CallMethod(static_cast<size_t>(MAXWELL3D_REG_INDEX(inline_data)),
                              regs.transform_feedback.controls[0].stride, true);
+
+        maxwell3d.Rasterizer().RegisterTransformFeedback(regs.upload.dest.Address());
     }
 };
 

src/video_core/rasterizer_interface.h

@@ -173,5 +173,13 @@ public:
     virtual void BindChannel(Tegra::Control::ChannelState& channel) {}
 
     virtual void ReleaseChannel(s32 channel_id) {}
+
+    /// Register the address as a Transform Feedback Object
+    virtual void RegisterTransformFeedback(GPUVAddr tfb_object_addr) {}
+
+    /// Returns true when the rasterizer has Draw Transform Feedback capabilities
+    virtual bool HasDrawTransformFeedback() {
+        return false;
+    }
 };
 } // namespace VideoCore

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -376,4 +376,15 @@ void BufferCacheRuntime::BindImageBuffer(Buffer& buffer, u32 offset, u32 size, P
     *image_handles++ = buffer.View(offset, size, format);
 }
 
+void BufferCacheRuntime::BindTransformFeedbackObject(GPUVAddr tfb_object_addr) {
+    OGLTransformFeedback& tfb_object = tfb_objects[tfb_object_addr];
+    tfb_object.Create();
+    glBindTransformFeedback(GL_TRANSFORM_FEEDBACK, tfb_object.handle);
+}
+
+GLuint BufferCacheRuntime::GetTransformFeedbackObject(GPUVAddr tfb_object_addr) {
+    ASSERT(tfb_objects.contains(tfb_object_addr));
+    return tfb_objects[tfb_object_addr].handle;
+}
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -5,6 +5,7 @@
 
 #include <array>
 #include <span>
+#include <unordered_map>
 
 #include "common/common_types.h"
 #include "video_core/buffer_cache/buffer_cache_base.h"
@@ -121,6 +122,9 @@ public:
     void BindImageBuffer(Buffer& buffer, u32 offset, u32 size,
                          VideoCore::Surface::PixelFormat format);
 
+    void BindTransformFeedbackObject(GPUVAddr tfb_object_addr);
+    GLuint GetTransformFeedbackObject(GPUVAddr tfb_object_addr);
+
     u64 GetDeviceMemoryUsage() const;
 
     void BindFastUniformBuffer(size_t stage, u32 binding_index, u32 size) {
@@ -233,6 +237,7 @@ private:
     u32 index_buffer_offset = 0;
 
     u64 device_access_memory;
+    std::unordered_map<GPUVAddr, OGLTransformFeedback> tfb_objects;
 };
 
 struct BufferCacheParams {

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -309,6 +309,13 @@ void RasterizerOpenGL::DrawIndirect() {
     const auto& params = maxwell3d->draw_manager->GetIndirectParams();
     buffer_cache.SetDrawIndirect(&params);
     PrepareDraw(params.is_indexed, [this, &params](GLenum primitive_mode) {
+        if (params.is_byte_count) {
+            const GPUVAddr tfb_object_base_addr = params.indirect_start_address - 4U;
+            const GLuint tfb_object =
+                buffer_cache_runtime.GetTransformFeedbackObject(tfb_object_base_addr);
+            glDrawTransformFeedback(primitive_mode, tfb_object);
+            return;
+        }
         const auto [buffer, offset] = buffer_cache.GetDrawIndirectBuffer();
         const GLvoid* const gl_offset =
             reinterpret_cast<const GLvoid*>(static_cast<uintptr_t>(offset));
@@ -1371,6 +1378,10 @@ void RasterizerOpenGL::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
+void RasterizerOpenGL::RegisterTransformFeedback(GPUVAddr tfb_object_addr) {
+    buffer_cache_runtime.BindTransformFeedbackObject(tfb_object_addr);
+}
+
 AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_, TextureCache& texture_cache_)
     : buffer_cache{buffer_cache_}, texture_cache{texture_cache_} {}
 

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -139,6 +139,12 @@ public:
 
     void ReleaseChannel(s32 channel_id) override;
 
+    void RegisterTransformFeedback(GPUVAddr tfb_object_addr) override;
+
+    bool HasDrawTransformFeedback() override {
+        return true;
+    }
+
 private:
     static constexpr size_t MAX_TEXTURES = 192;
     static constexpr size_t MAX_IMAGES = 48;

src/video_core/renderer_opengl/gl_resource_manager.cpp

@@ -207,4 +207,21 @@ void OGLQuery::Release() {
     handle = 0;
 }
 
+void OGLTransformFeedback::Create() {
+    if (handle != 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
+    glCreateTransformFeedbacks(1, &handle);
+}
+
+void OGLTransformFeedback::Release() {
+    if (handle == 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
+    glDeleteTransformFeedbacks(1, &handle);
+    handle = 0;
+}
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -323,4 +323,31 @@ public:
     GLuint handle = 0;
 };
 
+class OGLTransformFeedback final {
+public:
+    YUZU_NON_COPYABLE(OGLTransformFeedback);
+
+    OGLTransformFeedback() = default;
+
+    OGLTransformFeedback(OGLTransformFeedback&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+
+    ~OGLTransformFeedback() {
+        Release();
+    }
+
+    OGLTransformFeedback& operator=(OGLTransformFeedback&& o) noexcept {
+        Release();
+        handle = std::exchange(o.handle, 0);
+        return *this;
+    }
+
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create();
+
+    /// Deletes the internal OpenGL resource
+    void Release();
+
+    GLuint handle = 0;
+};
+
 } // namespace OpenGL

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -233,6 +233,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
           .ignore_nan_fp_comparisons = true,
           .gl_max_compute_smem_size = device.GetMaxComputeSharedMemorySize(),
           .min_ssbo_alignment = device.GetShaderStorageBufferAlignment(),
+          .max_user_clip_distances = 8,
       },
       host_info{
           .support_float64 = true,

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -374,6 +374,7 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
         .has_broken_robust =
             device.IsNvidia() && device.GetNvidiaArch() <= NvidiaArchitecture::Arch_Pascal,
         .min_ssbo_alignment = device.GetStorageBufferAlignment(),
+        .max_user_clip_distances = device.GetMaxUserClipDistances(),
     };
 
     host_info = Shader::HostTranslateInfo{

src/video_core/texture_cache/decode_bc.cpp

@@ -60,66 +60,72 @@ u32 ConvertedBytesPerBlock(VideoCore::Surface::PixelFormat pixel_format) {
 }
 
 template <auto decompress, PixelFormat pixel_format>
-void DecompressBlocks(std::span<const u8> input, std::span<u8> output, Extent3D extent,
+void DecompressBlocks(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
                       bool is_signed = false) {
     const u32 out_bpp = ConvertedBytesPerBlock(pixel_format);
-    const u32 block_width = std::min(extent.width, BLOCK_SIZE);
-    const u32 block_height = std::min(extent.height, BLOCK_SIZE);
-    const u32 pitch = extent.width * out_bpp;
+    const u32 block_size = BlockSize(pixel_format);
+    const u32 width = copy.image_extent.width;
+    const u32 height = copy.image_extent.height * copy.image_subresource.num_layers;
+    const u32 depth = copy.image_extent.depth;
+    const u32 block_width = std::min(width, BLOCK_SIZE);
+    const u32 block_height = std::min(height, BLOCK_SIZE);
+    const u32 pitch = width * out_bpp;
     size_t input_offset = 0;
     size_t output_offset = 0;
-    for (u32 slice = 0; slice < extent.depth; ++slice) {
-        for (u32 y = 0; y < extent.height; y += block_height) {
-            size_t row_offset = 0;
-            for (u32 x = 0; x < extent.width;
-                 x += block_width, row_offset += block_width * out_bpp) {
-                const u8* src = input.data() + input_offset;
-                u8* const dst = output.data() + output_offset + row_offset;
+    for (u32 slice = 0; slice < depth; ++slice) {
+        for (u32 y = 0; y < height; y += block_height) {
+            size_t src_offset = input_offset;
+            size_t dst_offset = output_offset;
+            for (u32 x = 0; x < width; x += block_width) {
+                const u8* src = input.data() + src_offset;
+                u8* const dst = output.data() + dst_offset;
                 if constexpr (IsSigned(pixel_format)) {
-                    decompress(src, dst, x, y, extent.width, extent.height, is_signed);
+                    decompress(src, dst, x, y, width, height, is_signed);
                 } else {
-                    decompress(src, dst, x, y, extent.width, extent.height);
+                    decompress(src, dst, x, y, width, height);
                 }
-                input_offset += BlockSize(pixel_format);
+                src_offset += block_size;
+                dst_offset += block_width * out_bpp;
             }
+            input_offset += copy.buffer_row_length * block_size / block_width;
             output_offset += block_height * pitch;
         }
     }
 }
 
-void DecompressBCn(std::span<const u8> input, std::span<u8> output, Extent3D extent,
+void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
                    VideoCore::Surface::PixelFormat pixel_format) {
     switch (pixel_format) {
     case PixelFormat::BC1_RGBA_UNORM:
     case PixelFormat::BC1_RGBA_SRGB:
-        DecompressBlocks<bcn::DecodeBc1, PixelFormat::BC1_RGBA_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc1, PixelFormat::BC1_RGBA_UNORM>(input, output, copy);
         break;
     case PixelFormat::BC2_UNORM:
     case PixelFormat::BC2_SRGB:
-        DecompressBlocks<bcn::DecodeBc2, PixelFormat::BC2_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc2, PixelFormat::BC2_UNORM>(input, output, copy);
         break;
     case PixelFormat::BC3_UNORM:
     case PixelFormat::BC3_SRGB:
-        DecompressBlocks<bcn::DecodeBc3, PixelFormat::BC3_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc3, PixelFormat::BC3_UNORM>(input, output, copy);
         break;
     case PixelFormat::BC4_SNORM:
     case PixelFormat::BC4_UNORM:
         DecompressBlocks<bcn::DecodeBc4, PixelFormat::BC4_UNORM>(
-            input, output, extent, pixel_format == PixelFormat::BC4_SNORM);
+            input, output, copy, pixel_format == PixelFormat::BC4_SNORM);
         break;
     case PixelFormat::BC5_SNORM:
     case PixelFormat::BC5_UNORM:
         DecompressBlocks<bcn::DecodeBc5, PixelFormat::BC5_UNORM>(
-            input, output, extent, pixel_format == PixelFormat::BC5_SNORM);
+            input, output, copy, pixel_format == PixelFormat::BC5_SNORM);
         break;
     case PixelFormat::BC6H_SFLOAT:
     case PixelFormat::BC6H_UFLOAT:
         DecompressBlocks<bcn::DecodeBc6, PixelFormat::BC6H_UFLOAT>(
-            input, output, extent, pixel_format == PixelFormat::BC6H_SFLOAT);
+            input, output, copy, pixel_format == PixelFormat::BC6H_SFLOAT);
         break;
     case PixelFormat::BC7_SRGB:
     case PixelFormat::BC7_UNORM:
-        DecompressBlocks<bcn::DecodeBc7, PixelFormat::BC7_UNORM>(input, output, extent);
+        DecompressBlocks<bcn::DecodeBc7, PixelFormat::BC7_UNORM>(input, output, copy);
         break;
     default:
         LOG_WARNING(HW_GPU, "Unimplemented BCn decompression {}", pixel_format);

src/video_core/texture_cache/decode_bc.h

@@ -13,7 +13,7 @@ namespace VideoCommon {
 
 [[nodiscard]] u32 ConvertedBytesPerBlock(VideoCore::Surface::PixelFormat pixel_format);
 
-void DecompressBCn(std::span<const u8> input, std::span<u8> output, Extent3D extent,
+void DecompressBCn(std::span<const u8> input, std::span<u8> output, BufferImageCopy& copy,
                    VideoCore::Surface::PixelFormat pixel_format);
 
 } // namespace VideoCommon

src/video_core/texture_cache/util.cpp

@@ -837,6 +837,7 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
                                                                    std::span<u8> output) {
     const size_t guest_size_bytes = input.size_bytes();
     const u32 bpp_log2 = BytesPerBlockLog2(info.format);
+    const Extent2D tile_size = DefaultBlockSize(info.format);
     const Extent3D size = info.size;
 
     if (info.type == ImageType::Linear) {
@@ -847,7 +848,7 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
         return {{
             .buffer_offset = 0,
             .buffer_size = guest_size_bytes,
-            .buffer_row_length = info.pitch >> bpp_log2,
+            .buffer_row_length = info.pitch * tile_size.width >> bpp_log2,
             .buffer_image_height = size.height,
             .image_subresource =
                 {
@@ -862,7 +863,6 @@ boost::container::small_vector<BufferImageCopy, 16> UnswizzleImage(Tegra::Memory
     const LevelInfo level_info = MakeLevelInfo(info);
     const s32 num_layers = info.resources.layers;
     const s32 num_levels = info.resources.levels;
-    const Extent2D tile_size = DefaultBlockSize(info.format);
     const std::array level_sizes = CalculateLevelSizes(level_info, num_levels);
     const Extent2D gob = GobSize(bpp_log2, info.block.height, info.tile_width_spacing);
     const u32 layer_size = CalculateLevelBytes(level_sizes, num_levels);
@@ -926,8 +926,6 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
 
         const auto input_offset = input.subspan(copy.buffer_offset);
         copy.buffer_offset = output_offset;
-        copy.buffer_row_length = mip_size.width;
-        copy.buffer_image_height = mip_size.height;
 
         const auto recompression_setting = Settings::values.astc_recompression.GetValue();
         const bool astc = IsPixelFormatASTC(info.format);
@@ -972,16 +970,14 @@ void ConvertImage(std::span<const u8> input, const ImageInfo& info, std::span<u8
                 bpp_div;
             output_offset += static_cast<u32>(copy.buffer_size);
         } else {
-            const Extent3D image_extent{
-                .width = copy.image_extent.width,
-                .height = copy.image_extent.height * copy.image_subresource.num_layers,
-                .depth = copy.image_extent.depth,
-            };
-            DecompressBCn(input_offset, output.subspan(output_offset), image_extent, info.format);
+            DecompressBCn(input_offset, output.subspan(output_offset), copy, info.format);
             output_offset += copy.image_extent.width * copy.image_extent.height *
                              copy.image_subresource.num_layers *
                              ConvertedBytesPerBlock(info.format);
         }
+
+        copy.buffer_row_length = mip_size.width;
+        copy.buffer_image_height = mip_size.height;
     }
 }
 

src/video_core/vulkan_common/vulkan_device.h

@@ -665,6 +665,10 @@ public:
         return properties.properties.limits.maxViewports;
     }
 
+    u32 GetMaxUserClipDistances() const {
+        return properties.properties.limits.maxClipDistances;
+    }
+
     bool SupportsConditionalBarriers() const {
         return supports_conditional_barriers;
     }

src/yuzu/main.h

@@ -168,14 +168,6 @@ class GMainWindow : public QMainWindow {
     /// Max number of recently loaded items to keep track of
     static const int max_recent_files_item = 10;
 
-    // TODO: Make use of this!
-    enum {
-        UI_IDLE,
-        UI_EMU_BOOTING,
-        UI_EMU_RUNNING,
-        UI_EMU_STOPPING,
-    };
-
     enum {
         CREATE_SHORTCUT_MSGBOX_FULLSCREEN_YES,
         CREATE_SHORTCUT_MSGBOX_SUCCESS,