SMMU: Initial adaptation to video_core.

src/audio_core/device/device_session.cpp

@@ -9,6 +9,7 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/memory.h"
+#include "core/guest_memory.h"
 
 #include "core/hle/kernel/k_process.h"
 

src/audio_core/renderer/command/data_source/decode.cpp

@@ -10,6 +10,7 @@
 #include "common/logging/log.h"
 #include "common/scratch_buffer.h"
 #include "core/memory.h"
+#include "core/guest_memory.h"
 
 namespace AudioCore::Renderer {
 

src/core/CMakeLists.txt

@@ -37,6 +37,8 @@ add_library(core STATIC
     debugger/gdbstub_arch.h
     debugger/gdbstub.cpp
     debugger/gdbstub.h
+    device_memory_manager.h
+    device_memory_manager.inc
     device_memory.cpp
     device_memory.h
     file_sys/fssystem/fs_i_storage.h

src/core/core.cpp

@@ -651,7 +651,7 @@ size_t System::GetCurrentHostThreadID() const {
     return impl->kernel.GetCurrentHostThreadID();
 }
 
-void System::GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback) {
+void System::GatherGPUDirtyMemory(std::function<void(PAddr, size_t)>& callback) {
     return this->ApplicationProcess()->GatherGPUDirtyMemory(callback);
 }
 

src/core/core.h

@@ -224,7 +224,7 @@ public:
     /// Prepare the core emulation for a reschedule
     void PrepareReschedule(u32 core_index);
 
-    void GatherGPUDirtyMemory(std::function<void(VAddr, size_t)>& callback);
+    void GatherGPUDirtyMemory(std::function<void(PAddr, size_t)>& callback);
 
     [[nodiscard]] size_t GetCurrentHostThreadID() const;
 

src/core/device_memory_manager.h

@@ -3,10 +3,11 @@
 
 #pragma once
 
-#include <deque>
-#include <memory>
 #include <array>
 #include <atomic>
+#include <deque>
+#include <memory>
+#include <mutex>
 
 #include "common/common_types.h"
 #include "common/virtual_buffer.h"
@@ -48,26 +49,54 @@ public:
     template <typename T>
     const T* GetPointer(DAddr address) const;
 
+    DAddr GetAddressFromPAddr(PAddr address) const {
+        DAddr subbits = static_cast<DAddr>(address & page_mask);
+        return (static_cast<DAddr>(compressed_device_addr[(address >> page_bits)]) << page_bits) + subbits;
+    }
+
+    PAddr GetPhysicalRawAddressFromDAddr(DAddr address) const {
+        PAddr subbits = static_cast<PAddr>(address & page_mask);
+        auto paddr = compressed_physical_ptr[(address >> page_bits)];
+        if (paddr == 0) {
+            return 0;
+        }
+        return (static_cast<PAddr>(paddr - 1) << page_bits) + subbits;
+    }
+
     template <typename T>
     void Write(DAddr address, T value);
 
     template <typename T>
     T Read(DAddr address) const;
 
+    const u8* GetSpan(const DAddr src_addr, const std::size_t size) const {
+        return nullptr;
+    }
+
+    u8* GetSpan(const DAddr src_addr, const std::size_t size) {
+        return nullptr;
+    }
+
     void ReadBlock(DAddr address, void* dest_pointer, size_t size);
-    void WriteBlock(DAddr address, void* src_pointer, size_t size);
+    void ReadBlockUnsafe(DAddr address, void* dest_pointer, size_t size);
+    void WriteBlock(DAddr address, const void* src_pointer, size_t size);
+    void WriteBlockUnsafe(DAddr address, const void* src_pointer, size_t size);
 
     size_t RegisterProcess(Memory::Memory* memory);
     void UnregisterProcess(size_t id);
 
     void UpdatePagesCachedCount(DAddr addr, size_t size, s32 delta);
 
+    static constexpr size_t AS_BITS = Traits::device_virtual_bits;
+
 private:
     static constexpr bool supports_pinning = Traits::supports_pinning;
     static constexpr size_t device_virtual_bits = Traits::device_virtual_bits;
     static constexpr size_t device_as_size = 1ULL << device_virtual_bits;
     static constexpr size_t physical_max_bits = 33;
     static constexpr size_t page_bits = 12;
+    static constexpr size_t page_size = 1ULL << page_bits;
+    static constexpr size_t page_mask = page_size - 1ULL;
     static constexpr u32 physical_address_base = 1U << page_bits;
 
     template <typename T>
@@ -136,11 +165,15 @@ private:
     private:
         std::array<std::atomic_uint16_t, subentries> values{};
     };
-    static_assert(sizeof(CounterEntry) == subentries * sizeof(u16), "CounterEntry should be 8 bytes!");
+    static_assert(sizeof(CounterEntry) == subentries * sizeof(u16),
+                  "CounterEntry should be 8 bytes!");
 
-    static constexpr size_t num_counter_entries = (1ULL << (device_virtual_bits - page_bits)) / subentries;
+    static constexpr size_t num_counter_entries =
+        (1ULL << (device_virtual_bits - page_bits)) / subentries;
     using CachedPages = std::array<CounterEntry, num_counter_entries>;
     std::unique_ptr<CachedPages> cached_pages;
+    std::mutex counter_guard;
+    std::mutex mapping_guard;
 };
 
 } // namespace Core

src/core/device_memory_manager.inc

@@ -105,7 +105,8 @@ template <typename Traits>
 DeviceMemoryManager<Traits>::DeviceMemoryManager(const DeviceMemory& device_memory_)
     : physical_base{reinterpret_cast<const uintptr_t>(device_memory_.buffer.BackingBasePointer())},
       interface{nullptr}, compressed_physical_ptr(device_as_size >> Memory::YUZU_PAGEBITS),
-      compressed_device_addr(1ULL << (physical_max_bits - Memory::YUZU_PAGEBITS)) {
+      compressed_device_addr(1ULL << (physical_max_bits - Memory::YUZU_PAGEBITS)),
+      cpu_backing_address(device_as_size >> Memory::YUZU_PAGEBITS) {
     impl = std::make_unique<DeviceMemoryManagerAllocator<Traits>>();
     cached_pages = std::make_unique<CachedPages>();
 }
@@ -144,10 +145,10 @@ void DeviceMemoryManager<Traits>::Map(DAddr address, VAddr virtual_address, size
     Core::Memory::Memory* process_memory = registered_processes[process_id];
     size_t start_page_d = address >> Memory::YUZU_PAGEBITS;
     size_t num_pages = Common::AlignUp(size, Memory::YUZU_PAGESIZE) >> Memory::YUZU_PAGEBITS;
-    std::atomic_thread_fence(std::memory_order_acquire);
+    std::scoped_lock lk(mapping_guard);
     for (size_t i = 0; i < num_pages; i++) {
         const VAddr new_vaddress = virtual_address + i * Memory::YUZU_PAGESIZE;
-        auto* ptr = process_memory->GetPointer(Common::ProcessAddress(new_vaddress));
+        auto* ptr = process_memory->GetPointerSilent(Common::ProcessAddress(new_vaddress));
         if (ptr == nullptr) [[unlikely]] {
             compressed_physical_ptr[start_page_d + i] = 0;
             continue;
@@ -157,14 +158,14 @@ void DeviceMemoryManager<Traits>::Map(DAddr address, VAddr virtual_address, size
         compressed_device_addr[phys_addr - 1U] = static_cast<u32>(start_page_d + i);
         InsertCPUBacking(start_page_d + i, new_vaddress, process_id);
     }
-    std::atomic_thread_fence(std::memory_order_release);
 }
 
 template <typename Traits>
 void DeviceMemoryManager<Traits>::Unmap(DAddr address, size_t size) {
     size_t start_page_d = address >> Memory::YUZU_PAGEBITS;
     size_t num_pages = Common::AlignUp(size, Memory::YUZU_PAGESIZE) >> Memory::YUZU_PAGEBITS;
-    std::atomic_thread_fence(std::memory_order_acquire);
+    interface->InvalidateRegion(address, size);
+    std::scoped_lock lk(mapping_guard);
     for (size_t i = 0; i < num_pages; i++) {
         auto phys_addr = compressed_physical_ptr[start_page_d + i];
         compressed_physical_ptr[start_page_d + i] = 0;
@@ -173,7 +174,6 @@ void DeviceMemoryManager<Traits>::Unmap(DAddr address, size_t size) {
             compressed_device_addr[phys_addr - 1] = 0;
         }
     }
-    std::atomic_thread_fence(std::memory_order_release);
 }
 
 template <typename Traits>
@@ -256,6 +256,7 @@ void DeviceMemoryManager<Traits>::WalkBlock(DAddr addr, std::size_t size, auto o
 
 template <typename Traits>
 void DeviceMemoryManager<Traits>::ReadBlock(DAddr address, void* dest_pointer, size_t size) {
+    interface->FlushRegion(address, size);
     WalkBlock(
         address, size,
         [&](size_t copy_amount, DAddr current_vaddr) {
@@ -274,7 +275,7 @@ void DeviceMemoryManager<Traits>::ReadBlock(DAddr address, void* dest_pointer, s
 }
 
 template <typename Traits>
-void DeviceMemoryManager<Traits>::WriteBlock(DAddr address, void* src_pointer, size_t size) {
+void DeviceMemoryManager<Traits>::WriteBlock(DAddr address, const void* src_pointer, size_t size) {
     WalkBlock(
         address, size,
         [&](size_t copy_amount, DAddr current_vaddr) {
@@ -287,7 +288,46 @@ void DeviceMemoryManager<Traits>::WriteBlock(DAddr address, void* src_pointer, s
             std::memcpy(dst_ptr, src_pointer, copy_amount);
         },
         [&](const std::size_t copy_amount) {
-            src_pointer = static_cast<u8*>(src_pointer) + copy_amount;
+            src_pointer = static_cast<const u8*>(src_pointer) + copy_amount;
+        });
+    interface->InvalidateRegion(address, size);
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::ReadBlockUnsafe(DAddr address, void* dest_pointer, size_t size) {
+    WalkBlock(
+        address, size,
+        [&](size_t copy_amount, DAddr current_vaddr) {
+            LOG_ERROR(
+                HW_Memory,
+                "Unmapped Device ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                current_vaddr, address, size);
+            std::memset(dest_pointer, 0, copy_amount);
+        },
+        [&](size_t copy_amount, const u8* const src_ptr) {
+            std::memcpy(dest_pointer, src_ptr, copy_amount);
+        },
+        [&](const std::size_t copy_amount) {
+            dest_pointer = static_cast<u8*>(dest_pointer) + copy_amount;
+        });
+}
+
+template <typename Traits>
+void DeviceMemoryManager<Traits>::WriteBlockUnsafe(DAddr address, const void* src_pointer,
+                                                   size_t size) {
+    WalkBlock(
+        address, size,
+        [&](size_t copy_amount, DAddr current_vaddr) {
+            LOG_ERROR(
+                HW_Memory,
+                "Unmapped Device WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
+                current_vaddr, address, size);
+        },
+        [&](size_t copy_amount, u8* const dst_ptr) {
+            std::memcpy(dst_ptr, src_pointer, copy_amount);
+        },
+        [&](const std::size_t copy_amount) {
+            src_pointer = static_cast<const u8*>(src_pointer) + copy_amount;
         });
 }
 
@@ -313,6 +353,18 @@ void DeviceMemoryManager<Traits>::UnregisterProcess(size_t id) {
 
 template <typename Traits>
 void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size, s32 delta) {
+    bool locked = false;
+    auto lock = [&] {
+        if (!locked) {
+            counter_guard.lock();
+            locked = true;
+        }
+    };
+    SCOPE_EXIT({
+        if (locked) {
+            counter_guard.unlock();
+        }
+    });
     u64 uncache_begin = 0;
     u64 cache_begin = 0;
     u64 uncache_bytes = 0;
@@ -347,6 +399,7 @@ void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size
             }
             uncache_bytes += Memory::YUZU_PAGESIZE;
         } else if (uncache_bytes > 0) {
+            lock();
             MarkRegionCaching(memory_interface, uncache_begin << Memory::YUZU_PAGEBITS,
                               uncache_bytes, false);
             uncache_bytes = 0;
@@ -357,6 +410,7 @@ void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size
             }
             cache_bytes += Memory::YUZU_PAGESIZE;
         } else if (cache_bytes > 0) {
+            lock();
             MarkRegionCaching(memory_interface, cache_begin << Memory::YUZU_PAGEBITS, cache_bytes,
                               true);
             cache_bytes = 0;
@@ -364,10 +418,12 @@ void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size
         vpage++;
     }
     if (uncache_bytes > 0) {
+        lock();
         MarkRegionCaching(memory_interface, uncache_begin << Memory::YUZU_PAGEBITS, uncache_bytes,
                           false);
     }
     if (cache_bytes > 0) {
+        lock();
         MarkRegionCaching(memory_interface, cache_begin << Memory::YUZU_PAGEBITS, cache_bytes,
                           true);
     }

src/core/gpu_dirty_memory_manager.h

@@ -23,7 +23,7 @@ public:
 
     ~GPUDirtyMemoryManager() = default;
 
-    void Collect(VAddr address, size_t size) {
+    void Collect(PAddr address, size_t size) {
         TransformAddress t = BuildTransform(address, size);
         TransformAddress tmp, original;
         do {
@@ -47,7 +47,7 @@ public:
                                                 std::memory_order_relaxed));
     }
 
-    void Gather(std::function<void(VAddr, size_t)>& callback) {
+    void Gather(std::function<void(PAddr, size_t)>& callback) {
         {
             std::scoped_lock lk(guard);
             TransformAddress t = current.exchange(default_transform, std::memory_order_relaxed);
@@ -65,7 +65,7 @@ public:
                 mask = mask >> empty_bits;
 
                 const size_t continuous_bits = std::countr_one(mask);
-                callback((static_cast<VAddr>(transform.address) << page_bits) + offset,
+                callback((static_cast<PAddr>(transform.address) << page_bits) + offset,
                          continuous_bits << align_bits);
                 mask = continuous_bits < align_size ? (mask >> continuous_bits) : 0;
                 offset += continuous_bits << align_bits;
@@ -89,7 +89,7 @@ private:
     constexpr static size_t align_mask = align_size - 1;
     constexpr static TransformAddress default_transform = {.address = ~0U, .mask = 0U};
 
-    bool IsValid(VAddr address) {
+    bool IsValid(PAddr address) {
         return address < (1ULL << 39);
     }
 
@@ -103,7 +103,7 @@ private:
         return mask;
     }
 
-    TransformAddress BuildTransform(VAddr address, size_t size) {
+    TransformAddress BuildTransform(PAddr address, size_t size) {
         const size_t minor_address = address & page_mask;
         const size_t minor_bit = minor_address >> align_bits;
         const size_t top_bit = (minor_address + size + align_mask) >> align_bits;

src/core/guest_memory.h

@@ -0,0 +1,218 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <iterator>
+#include <memory>
+#include <optional>
+#include <span>
+#include <vector>
+
+#include "common/scratch_buffer.h"
+#include "core/memory.h"
+
+namespace Core::Memory {
+
+enum GuestMemoryFlags : u32 {
+    Read = 1 << 0,
+    Write = 1 << 1,
+    Safe = 1 << 2,
+    Cached = 1 << 3,
+
+    SafeRead = Read | Safe,
+    SafeWrite = Write | Safe,
+    SafeReadWrite = SafeRead | SafeWrite,
+    SafeReadCachedWrite = SafeReadWrite | Cached,
+
+    UnsafeRead = Read,
+    UnsafeWrite = Write,
+    UnsafeReadWrite = UnsafeRead | UnsafeWrite,
+    UnsafeReadCachedWrite = UnsafeReadWrite | Cached,
+};
+
+namespace {
+template <typename M, typename T, GuestMemoryFlags FLAGS>
+class GuestMemory {
+    using iterator = T*;
+    using const_iterator = const T*;
+    using value_type = T;
+    using element_type = T;
+    using iterator_category = std::contiguous_iterator_tag;
+
+public:
+    GuestMemory() = delete;
+    explicit GuestMemory(M& memory, u64 addr, std::size_t size,
+                         Common::ScratchBuffer<T>* backup = nullptr)
+        : m_memory{memory}, m_addr{addr}, m_size{size} {
+        static_assert(FLAGS & GuestMemoryFlags::Read || FLAGS & GuestMemoryFlags::Write);
+        if constexpr (FLAGS & GuestMemoryFlags::Read) {
+            Read(addr, size, backup);
+        }
+    }
+
+    ~GuestMemory() = default;
+
+    T* data() noexcept {
+        return m_data_span.data();
+    }
+
+    const T* data() const noexcept {
+        return m_data_span.data();
+    }
+
+    size_t size() const noexcept {
+        return m_size;
+    }
+
+    size_t size_bytes() const noexcept {
+        return this->size() * sizeof(T);
+    }
+
+    [[nodiscard]] T* begin() noexcept {
+        return this->data();
+    }
+
+    [[nodiscard]] const T* begin() const noexcept {
+        return this->data();
+    }
+
+    [[nodiscard]] T* end() noexcept {
+        return this->data() + this->size();
+    }
+
+    [[nodiscard]] const T* end() const noexcept {
+        return this->data() + this->size();
+    }
+
+    T& operator[](size_t index) noexcept {
+        return m_data_span[index];
+    }
+
+    const T& operator[](size_t index) const noexcept {
+        return m_data_span[index];
+    }
+
+    void SetAddressAndSize(u64 addr, std::size_t size) noexcept {
+        m_addr = addr;
+        m_size = size;
+        m_addr_changed = true;
+    }
+
+    std::span<T> Read(u64 addr, std::size_t size,
+                      Common::ScratchBuffer<T>* backup = nullptr) noexcept {
+        m_addr = addr;
+        m_size = size;
+        if (m_size == 0) {
+            m_is_data_copy = true;
+            return {};
+        }
+
+        if (this->TrySetSpan()) {
+            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+                m_memory.FlushRegion(m_addr, this->size_bytes());
+            }
+        } else {
+            if (backup) {
+                backup->resize_destructive(this->size());
+                m_data_span = *backup;
+            } else {
+                m_data_copy.resize(this->size());
+                m_data_span = std::span(m_data_copy);
+            }
+            m_is_data_copy = true;
+            m_span_valid = true;
+            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+                m_memory.ReadBlock(m_addr, this->data(), this->size_bytes());
+            } else {
+                m_memory.ReadBlockUnsafe(m_addr, this->data(), this->size_bytes());
+            }
+        }
+        return m_data_span;
+    }
+
+    void Write(std::span<T> write_data) noexcept {
+        if constexpr (FLAGS & GuestMemoryFlags::Cached) {
+            m_memory.WriteBlockCached(m_addr, write_data.data(), this->size_bytes());
+        } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+            m_memory.WriteBlock(m_addr, write_data.data(), this->size_bytes());
+        } else {
+            m_memory.WriteBlockUnsafe(m_addr, write_data.data(), this->size_bytes());
+        }
+    }
+
+    bool TrySetSpan() noexcept {
+        if (u8* ptr = m_memory.GetSpan(m_addr, this->size_bytes()); ptr) {
+            m_data_span = {reinterpret_cast<T*>(ptr), this->size()};
+            m_span_valid = true;
+            return true;
+        }
+        return false;
+    }
+
+protected:
+    bool IsDataCopy() const noexcept {
+        return m_is_data_copy;
+    }
+
+    bool AddressChanged() const noexcept {
+        return m_addr_changed;
+    }
+
+    M& m_memory;
+    u64 m_addr{};
+    size_t m_size{};
+    std::span<T> m_data_span{};
+    std::vector<T> m_data_copy{};
+    bool m_span_valid{false};
+    bool m_is_data_copy{false};
+    bool m_addr_changed{false};
+};
+
+template <typename M, typename T, GuestMemoryFlags FLAGS>
+class GuestMemoryScoped : public GuestMemory<M, T, FLAGS> {
+public:
+    GuestMemoryScoped() = delete;
+    explicit GuestMemoryScoped(M& memory, u64 addr, std::size_t size,
+                               Common::ScratchBuffer<T>* backup = nullptr)
+        : GuestMemory<M, T, FLAGS>(memory, addr, size, backup) {
+        if constexpr (!(FLAGS & GuestMemoryFlags::Read)) {
+            if (!this->TrySetSpan()) {
+                if (backup) {
+                    this->m_data_span = *backup;
+                    this->m_span_valid = true;
+                    this->m_is_data_copy = true;
+                }
+            }
+        }
+    }
+
+    ~GuestMemoryScoped() {
+        if constexpr (FLAGS & GuestMemoryFlags::Write) {
+            if (this->size() == 0) [[unlikely]] {
+                return;
+            }
+
+            if (this->AddressChanged() || this->IsDataCopy()) {
+                ASSERT(this->m_span_valid);
+                if constexpr (FLAGS & GuestMemoryFlags::Cached) {
+                    this->m_memory.WriteBlockCached(this->m_addr, this->data(), this->size_bytes());
+                } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
+                    this->m_memory.WriteBlock(this->m_addr, this->data(), this->size_bytes());
+                } else {
+                    this->m_memory.WriteBlockUnsafe(this->m_addr, this->data(), this->size_bytes());
+                }
+            } else if constexpr ((FLAGS & GuestMemoryFlags::Safe) || (FLAGS & GuestMemoryFlags::Cached))  {
+                this->m_memory.InvalidateRegion(this->m_addr, this->size_bytes());
+            }
+        }
+    }
+};
+} // namespace
+
+template <typename T, GuestMemoryFlags FLAGS>
+using CpuGuestMemory = GuestMemory<Core::Memory::Memory, T, FLAGS>;
+template <typename T, GuestMemoryFlags FLAGS>
+using CpuGuestMemoryScoped = GuestMemoryScoped<Core::Memory::Memory, T, FLAGS>;
+
+} // namespace Tegra::Memory

src/core/hle/service/hle_ipc.cpp

@@ -22,19 +22,7 @@
 #include "core/hle/service/hle_ipc.h"
 #include "core/hle/service/ipc_helpers.h"
 #include "core/memory.h"
-
+#include "core/guest_memory.h"
-namespace {
-static thread_local std::array read_buffer_data_a{
-    Common::ScratchBuffer<u8>(),
-    Common::ScratchBuffer<u8>(),
-    Common::ScratchBuffer<u8>(),
-};
-static thread_local std::array read_buffer_data_x{
-    Common::ScratchBuffer<u8>(),
-    Common::ScratchBuffer<u8>(),
-    Common::ScratchBuffer<u8>(),
-};
-} // Anonymous namespace
 
 namespace Service {
 
@@ -343,48 +331,27 @@ std::vector<u8> HLERequestContext::ReadBufferCopy(std::size_t buffer_index) cons
 }
 
 std::span<const u8> HLERequestContext::ReadBufferA(std::size_t buffer_index) const {
-    static thread_local std::array read_buffer_a{
+    Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> gm(memory, 0, 0);
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-    };
 
     ASSERT_OR_EXECUTE_MSG(
         BufferDescriptorA().size() > buffer_index, { return {}; },
         "BufferDescriptorA invalid buffer_index {}", buffer_index);
-    auto& read_buffer = read_buffer_a[buffer_index];
+    return gm.Read(BufferDescriptorA()[buffer_index].Address(),
-    return read_buffer.Read(BufferDescriptorA()[buffer_index].Address(),
+                   BufferDescriptorA()[buffer_index].Size(), &read_buffer_data_a[buffer_index]);
-                            BufferDescriptorA()[buffer_index].Size(),
-                            &read_buffer_data_a[buffer_index]);
 }
 
 std::span<const u8> HLERequestContext::ReadBufferX(std::size_t buffer_index) const {
-    static thread_local std::array read_buffer_x{
+    Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> gm(memory, 0, 0);
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-    };
 
     ASSERT_OR_EXECUTE_MSG(
         BufferDescriptorX().size() > buffer_index, { return {}; },
         "BufferDescriptorX invalid buffer_index {}", buffer_index);
-    auto& read_buffer = read_buffer_x[buffer_index];
+    return gm.Read(BufferDescriptorX()[buffer_index].Address(),
-    return read_buffer.Read(BufferDescriptorX()[buffer_index].Address(),
+                   BufferDescriptorX()[buffer_index].Size(), &read_buffer_data_x[buffer_index]);
-                            BufferDescriptorX()[buffer_index].Size(),
-                            &read_buffer_data_x[buffer_index]);
 }
 
 std::span<const u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) const {
-    static thread_local std::array read_buffer_a{
+    Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> gm(memory, 0, 0);
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-    };
-    static thread_local std::array read_buffer_x{
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-        Core::Memory::CpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead>(memory, 0, 0),
-    };
 
     const bool is_buffer_a{BufferDescriptorA().size() > buffer_index &&
                            BufferDescriptorA()[buffer_index].Size()};
@@ -401,18 +368,14 @@ std::span<const u8> HLERequestContext::ReadBuffer(std::size_t buffer_index) cons
         ASSERT_OR_EXECUTE_MSG(
             BufferDescriptorA().size() > buffer_index, { return {}; },
             "BufferDescriptorA invalid buffer_index {}", buffer_index);
-        auto& read_buffer = read_buffer_a[buffer_index];
+        return gm.Read(BufferDescriptorA()[buffer_index].Address(),
-        return read_buffer.Read(BufferDescriptorA()[buffer_index].Address(),
+                       BufferDescriptorA()[buffer_index].Size(), &read_buffer_data_a[buffer_index]);
-                                BufferDescriptorA()[buffer_index].Size(),
-                                &read_buffer_data_a[buffer_index]);
     } else {
         ASSERT_OR_EXECUTE_MSG(
             BufferDescriptorX().size() > buffer_index, { return {}; },
             "BufferDescriptorX invalid buffer_index {}", buffer_index);
-        auto& read_buffer = read_buffer_x[buffer_index];
+        return gm.Read(BufferDescriptorX()[buffer_index].Address(),
-        return read_buffer.Read(BufferDescriptorX()[buffer_index].Address(),
+                       BufferDescriptorX()[buffer_index].Size(), &read_buffer_data_x[buffer_index]);
-                                BufferDescriptorX()[buffer_index].Size(),
-                                &read_buffer_data_x[buffer_index]);
     }
 }
 

src/core/hle/service/hle_ipc.h

@@ -19,8 +19,6 @@
 #include "core/hle/ipc.h"
 #include "core/hle/kernel/k_handle_table.h"
 #include "core/hle/kernel/svc_common.h"
-#include "core/hle/kernel/k_auto_object.h"
-#include "core/hle/kernel/k_handle_table.h"
 
 union Result;
 
@@ -377,10 +375,6 @@ public:
         return nullptr;
     }
 
-    Kernel::KScopedAutoObject<Kernel::KAutoObject> GetObjectFromHandle(u32 handle) {
-        return GetClientHandleTable().GetObjectForIpc(handle, thread);
-    }
-
     [[nodiscard]] std::shared_ptr<SessionRequestManager> GetManager() const {
         return manager.lock();
     }
@@ -432,6 +426,9 @@ private:
 
     Kernel::KernelCore& kernel;
     Core::Memory::Memory& memory;
+
+    mutable std::array<Common::ScratchBuffer<u8>, 3> read_buffer_data_a{};
+    mutable std::array<Common::ScratchBuffer<u8>, 3> read_buffer_data_x{};
 };
 
 } // namespace Service

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

@@ -2,6 +2,8 @@
 // SPDX-FileCopyrightText: 2022 Skyline Team and Contributors
 // SPDX-License-Identifier: GPL-3.0-or-later
 
+#include <functional>
+
 #include "common/alignment.h"
 #include "common/assert.h"
 #include "common/logging/log.h"
@@ -18,6 +20,7 @@ NvMap::Handle::Handle(u64 size_, Id id_)
 }
 
 NvResult NvMap::Handle::Alloc(Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress) {
+    std::scoped_lock lock(mutex);
     // Handles cannot be allocated twice
     if (allocated) {
         return NvResult::AccessDenied;
@@ -78,11 +81,9 @@ void NvMap::UnmapHandle(Handle& handle_description) {
 
     // Free and unmap the handle from the SMMU
     auto& smmu = host1x.MemoryManager();
-    smmu.Unmap(static_cast<DAddr>(handle_description.pin_virt_address),
+    smmu.Unmap(handle_description.d_address, handle_description.aligned_size);
-               handle_description.aligned_size);
+    smmu.Free(handle_description.d_address, static_cast<size_t>(handle_description.aligned_size));
-    smmu.Free(handle_description.pin_virt_address,
+    handle_description.d_address = 0;
-              static_cast<size_t>(handle_description.aligned_size));
-    handle_description.pin_virt_address = 0;
 }
 
 bool NvMap::TryRemoveHandle(const Handle& handle_description) {
@@ -123,41 +124,16 @@ std::shared_ptr<NvMap::Handle> NvMap::GetHandle(Handle::Id handle) {
     }
 }
 
-VAddr NvMap::GetHandleAddress(Handle::Id handle) {
+DAddr NvMap::GetHandleAddress(Handle::Id handle) {
     std::scoped_lock lock(handles_lock);
     try {
-        return handles.at(handle)->address;
+        return handles.at(handle)->d_address;
     } catch (std::out_of_range&) {
         return 0;
     }
 }
 
-NvResult NvMap::AllocateHandle(Handle::Id handle, Handle::Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress, size_t session_id) {
+DAddr NvMap::PinHandle(NvMap::Handle::Id handle, size_t session_id, bool low_area_pin) {
-    auto handle_description{GetHandle(handle)};
-    if (!handle_description) [[unlikely]] {
-        return NvResult::BadParameter;
-    }
-
-    if (handle_description->allocated) [[unlikely]] {
-        return NvResult::InsufficientMemory;
-    }
-
-    std::scoped_lock lock(handle_description->mutex);
-    NvResult result = handle_description->Alloc(pFlags, pAlign, pKind, pAddress);
-    if (result != NvResult::Success) {
-        return result;
-    }
-    auto& smmu = host1x.MemoryManager();
-    size_t total_size = static_cast<size_t>(handle_description->aligned_size);
-    handle_description->d_address = smmu.Allocate(total_size);
-    if (handle_description->d_address == 0) {
-        return NvResult::InsufficientMemory;
-    }
-    smmu.Map(handle_description->d_address, handle_description->address, total_size, session_id);
-    return NvResult::Success;
-}
-
-u32 NvMap::PinHandle(NvMap::Handle::Id handle, size_t session_id) {
     auto handle_description{GetHandle(handle)};
     if (!handle_description) [[unlikely]] {
         return 0;
@@ -176,35 +152,38 @@ u32 NvMap::PinHandle(NvMap::Handle::Id handle, size_t session_id) {
                 handle_description->unmap_queue_entry.reset();
 
                 handle_description->pins++;
-                return handle_description->pin_virt_address;
+                return handle_description->d_address;
             }
         }
 
+        using namespace std::placeholders;
         // If not then allocate some space and map it
         DAddr address{};
         auto& smmu = host1x.MemoryManager();
-        while ((address = smmu.AllocatePinned(
+        auto allocate = std::bind(&Tegra::MaxwellDeviceMemoryManager::Allocate, &smmu, _1);
-                    static_cast<size_t>(handle_description->aligned_size))) == 0) {
+                         //: std::bind(&Tegra::MaxwellDeviceMemoryManager::Allocate, &smmu, _1);
+        while ((address = allocate(static_cast<size_t>(handle_description->aligned_size))) == 0) {
             // Free handles until the allocation succeeds
             std::scoped_lock queueLock(unmap_queue_lock);
             if (auto freeHandleDesc{unmap_queue.front()}) {
                 // Handles in the unmap queue are guaranteed not to be pinned so don't bother
                 // checking if they are before unmapping
                 std::scoped_lock freeLock(freeHandleDesc->mutex);
-                if (handle_description->pin_virt_address)
+                if (handle_description->d_address)
                     UnmapHandle(*freeHandleDesc);
             } else {
                 LOG_CRITICAL(Service_NVDRV, "Ran out of SMMU address space!");
             }
         }
 
+        handle_description->d_address = address;
+
         smmu.Map(address, handle_description->address, handle_description->aligned_size,
                  session_id);
-        handle_description->pin_virt_address = static_cast<u32>(address);
     }
 
     handle_description->pins++;
-    return handle_description->pin_virt_address;
+    return handle_description->d_address;
 }
 
 void NvMap::UnpinHandle(Handle::Id handle) {
@@ -255,15 +234,10 @@ std::optional<NvMap::FreeInfo> NvMap::FreeHandle(Handle::Id handle, bool interna
                 LOG_WARNING(Service_NVDRV, "User duplicate count imbalance detected!");
             } else if (handle_description->dupes == 0) {
                 // Force unmap the handle
-                if (handle_description->pin_virt_address) {
+                if (handle_description->d_address) {
                     std::scoped_lock queueLock(unmap_queue_lock);
                     UnmapHandle(*handle_description);
                 }
-                if (handle_description->allocated) {
-                    auto& smmu = host1x.MemoryManager();
-                    smmu.Free(handle_description->d_address, handle_description->aligned_size);
-                    smmu.Unmap(handle_description->d_address, handle_description->aligned_size);
-                }
 
                 handle_description->pins = 0;
             }

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

@@ -48,7 +48,7 @@ public:
         using Id = u32;
         Id id; //!< A globally unique identifier for this handle
 
-        s32 pins{};
+        s64 pins{};
         u32 pin_virt_address{};
         std::optional<typename std::list<std::shared_ptr<Handle>>::iterator> unmap_queue_entry{};
 
@@ -63,15 +63,14 @@ public:
 
         VAddr address{};   //!< The memory location in the guest's AS that this handle corresponds to,
                            //!< this can also be in the nvdrv tmem
-        DAddr d_address{}; //!< The memory location in the device's AS that this handle corresponds to,
-                           //!< this can also be in the nvdrv tmem
         bool is_shared_mem_mapped{}; //!< If this nvmap has been mapped with the MapSharedMem IPC
                                      //!< call
 
         u8 kind{};        //!< Used for memory compression
         bool allocated{}; //!< If the handle has been allocated with `Alloc`
 
-        u64 dma_map_addr{}; //! remove me after implementing pinning.
+        DAddr d_address{}; //!< The memory location in the device's AS that this handle corresponds to,
+                           //!< this can also be in the nvdrv tmem
 
         Handle(u64 size, Id id);
 
@@ -119,7 +118,7 @@ public:
 
     std::shared_ptr<Handle> GetHandle(Handle::Id handle);
 
-    VAddr GetHandleAddress(Handle::Id handle);
+    DAddr GetHandleAddress(Handle::Id handle);
 
     /**
      * @brief Maps a handle into the SMMU address space
@@ -127,15 +126,7 @@ public:
      * number of calls to `UnpinHandle`
      * @return The SMMU virtual address that the handle has been mapped to
      */
-    u32 PinHandle(Handle::Id handle, size_t session_id);
+    DAddr PinHandle(Handle::Id handle, size_t session_id, bool low_area_pin);
-
-    /**
-     * @brief Maps a handle into the SMMU address space
-     * @note This operation is refcounted, the number of calls to this must eventually match the
-     * number of calls to `UnpinHandle`
-     * @return The SMMU virtual address that the handle has been mapped to
-     */
-    NvResult AllocateHandle(Handle::Id handle, Handle::Flags pFlags, u32 pAlign, u8 pKind, u64 pAddress, size_t session_id);
 
     /**
      * @brief When this has been called an equal number of times to `PinHandle` for the supplied

src/core/hle/service/nvdrv/devices/nvdisp_disp0.cpp

@@ -42,7 +42,7 @@ void nvdisp_disp0::flip(u32 buffer_handle, u32 offset, android::PixelFormat form
                         u32 height, u32 stride, android::BufferTransformFlags transform,
                         const Common::Rectangle<int>& crop_rect,
                         std::array<Service::Nvidia::NvFence, 4>& fences, u32 num_fences) {
-    const VAddr addr = nvmap.GetHandleAddress(buffer_handle);
+    const DAddr addr = nvmap.GetHandleAddress(buffer_handle);
     LOG_TRACE(Service,
               "Drawing from address {:X} offset {:08X} Width {} Height {} Stride {} Format {}",
               addr, offset, width, height, stride, format);

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

@@ -40,15 +40,15 @@ NvResult nvhost_as_gpu::Ioctl1(DeviceFD fd, Ioctl command, std::span<const u8> i
         case 0x3:
             return WrapFixed(this, &nvhost_as_gpu::FreeSpace, input, output);
         case 0x5:
-            return WrapFixed(this, &nvhost_as_gpu::UnmapBuffer, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::UnmapBuffer, input, output, fd);
         case 0x6:
-            return WrapFixed(this, &nvhost_as_gpu::MapBufferEx, input, output);
+            return WrapFixed(this, &nvhost_as_gpu::MapBufferEx, input, output, fd);
         case 0x8:
             return WrapFixed(this, &nvhost_as_gpu::GetVARegions1, input, output);
         case 0x9:
             return WrapFixed(this, &nvhost_as_gpu::AllocAsEx, input, output);
         case 0x14:
-            return WrapVariable(this, &nvhost_as_gpu::Remap, input, output);
+            return WrapVariable(this, &nvhost_as_gpu::Remap, input, output, fd);
         default:
             break;
         }
@@ -86,8 +86,15 @@ NvResult nvhost_as_gpu::Ioctl3(DeviceFD fd, Ioctl command, std::span<const u8> i
     return NvResult::NotImplemented;
 }
 
-void nvhost_as_gpu::OnOpen(size_t session_id, DeviceFD fd) {}
+void nvhost_as_gpu::OnOpen(size_t session_id, DeviceFD fd) {
-void nvhost_as_gpu::OnClose(DeviceFD fd) {}
+    sessions[fd] = session_id;
+}
+void nvhost_as_gpu::OnClose(DeviceFD fd) {
+    auto it = sessions.find(fd);
+    if (it != sessions.end()) {
+        sessions.erase(it);
+    }
+}
 
 NvResult nvhost_as_gpu::AllocAsEx(IoctlAllocAsEx& params) {
     LOG_DEBUG(Service_NVDRV, "called, big_page_size=0x{:X}", params.big_page_size);
@@ -206,6 +213,8 @@ void nvhost_as_gpu::FreeMappingLocked(u64 offset) {
                        static_cast<u32>(aligned_size >> page_size_bits));
     }
 
+    nvmap.UnpinHandle(mapping->handle);
+
     // Sparse mappings shouldn't be fully unmapped, just returned to their sparse state
     // Only FreeSpace can unmap them fully
     if (mapping->sparse_alloc) {
@@ -259,7 +268,7 @@ NvResult nvhost_as_gpu::FreeSpace(IoctlFreeSpace& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::Remap(std::span<IoctlRemapEntry> entries) {
+NvResult nvhost_as_gpu::Remap(std::span<IoctlRemapEntry> entries, DeviceFD fd) {
     LOG_DEBUG(Service_NVDRV, "called, num_entries=0x{:X}", entries.size());
 
     if (!vm.initialised) {
@@ -293,19 +302,19 @@ NvResult nvhost_as_gpu::Remap(std::span<IoctlRemapEntry> entries) {
                 return NvResult::BadValue;
             }
 
-            VAddr cpu_address{static_cast<VAddr>(
+            DAddr base = nvmap.PinHandle(entry.handle, sessions[fd], false);
-                handle->address +
+            DAddr device_address{static_cast<DAddr>(
-                (static_cast<u64>(entry.handle_offset_big_pages) << vm.big_page_size_bits))};
+                base + (static_cast<u64>(entry.handle_offset_big_pages) << vm.big_page_size_bits))};
 
-            gmmu->Map(virtual_address, cpu_address, size, static_cast<Tegra::PTEKind>(entry.kind),
+            gmmu->Map(virtual_address, device_address, size,
-                      use_big_pages);
+                      static_cast<Tegra::PTEKind>(entry.kind), use_big_pages);
         }
     }
 
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
+NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params, DeviceFD fd) {
     LOG_DEBUG(Service_NVDRV,
               "called, flags={:X}, nvmap_handle={:X}, buffer_offset={}, mapping_size={}"
               ", offset={}",
@@ -331,9 +340,9 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
             }
 
             u64 gpu_address{static_cast<u64>(params.offset + params.buffer_offset)};
-            VAddr cpu_address{mapping->ptr + params.buffer_offset};
+            VAddr device_address{mapping->ptr + params.buffer_offset};
 
-            gmmu->Map(gpu_address, cpu_address, params.mapping_size,
+            gmmu->Map(gpu_address, device_address, params.mapping_size,
                       static_cast<Tegra::PTEKind>(params.kind), mapping->big_page);
 
             return NvResult::Success;
@@ -349,7 +358,8 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
         return NvResult::BadValue;
     }
 
-    VAddr cpu_address{static_cast<VAddr>(handle->address + params.buffer_offset)};
+    DAddr device_address{static_cast<DAddr>(nvmap.PinHandle(params.handle, sessions[fd], false) +
+                                            params.buffer_offset)};
     u64 size{params.mapping_size ? params.mapping_size : handle->orig_size};
 
     bool big_page{[&]() {
@@ -373,15 +383,14 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
         }
 
         const bool use_big_pages = alloc->second.big_pages && big_page;
-        gmmu->Map(params.offset, cpu_address, size, static_cast<Tegra::PTEKind>(params.kind),
+        gmmu->Map(params.offset, device_address, size, static_cast<Tegra::PTEKind>(params.kind),
                   use_big_pages);
 
-        auto mapping{std::make_shared<Mapping>(cpu_address, params.offset, size, true,
+        auto mapping{std::make_shared<Mapping>(params.handle, device_address, params.offset, size,
-                                               use_big_pages, alloc->second.sparse)};
+                                               true, use_big_pages, alloc->second.sparse)};
         alloc->second.mappings.push_back(mapping);
         mapping_map[params.offset] = mapping;
     } else {
-
         auto& allocator{big_page ? *vm.big_page_allocator : *vm.small_page_allocator};
         u32 page_size{big_page ? vm.big_page_size : VM::YUZU_PAGESIZE};
         u32 page_size_bits{big_page ? vm.big_page_size_bits : VM::PAGE_SIZE_BITS};
@@ -394,18 +403,18 @@ NvResult nvhost_as_gpu::MapBufferEx(IoctlMapBufferEx& params) {
             return NvResult::InsufficientMemory;
         }
 
-        gmmu->Map(params.offset, cpu_address, Common::AlignUp(size, page_size),
+        gmmu->Map(params.offset, device_address, Common::AlignUp(size, page_size),
                   static_cast<Tegra::PTEKind>(params.kind), big_page);
 
-        auto mapping{
+        auto mapping{std::make_shared<Mapping>(params.handle, device_address, params.offset, size,
-            std::make_shared<Mapping>(cpu_address, params.offset, size, false, big_page, false)};
+                                               false, big_page, false)};
         mapping_map[params.offset] = mapping;
     }
 
     return NvResult::Success;
 }
 
-NvResult nvhost_as_gpu::UnmapBuffer(IoctlUnmapBuffer& params) {
+NvResult nvhost_as_gpu::UnmapBuffer(IoctlUnmapBuffer& params, DeviceFD fd) {
     LOG_DEBUG(Service_NVDRV, "called, offset=0x{:X}", params.offset);
 
     std::scoped_lock lock(mutex);
@@ -433,6 +442,8 @@ NvResult nvhost_as_gpu::UnmapBuffer(IoctlUnmapBuffer& params) {
             gmmu->Unmap(params.offset, mapping->size);
         }
 
+        nvmap.UnpinHandle(mapping->handle);
+
         mapping_map.erase(params.offset);
     } catch (const std::out_of_range&) {
         LOG_WARNING(Service_NVDRV, "Couldn't find region to unmap at 0x{:X}", params.offset);

src/core/hle/service/nvdrv/devices/nvhost_as_gpu.h

@@ -141,9 +141,9 @@ private:
 
     NvResult AllocAsEx(IoctlAllocAsEx& params);
     NvResult AllocateSpace(IoctlAllocSpace& params);
-    NvResult Remap(std::span<IoctlRemapEntry> params);
+    NvResult Remap(std::span<IoctlRemapEntry> params, DeviceFD fd);
-    NvResult MapBufferEx(IoctlMapBufferEx& params);
+    NvResult MapBufferEx(IoctlMapBufferEx& params, DeviceFD fd);
-    NvResult UnmapBuffer(IoctlUnmapBuffer& params);
+    NvResult UnmapBuffer(IoctlUnmapBuffer& params, DeviceFD fd);
     NvResult FreeSpace(IoctlFreeSpace& params);
     NvResult BindChannel(IoctlBindChannel& params);
 
@@ -159,16 +159,18 @@ private:
     NvCore::NvMap& nvmap;
 
     struct Mapping {
-        VAddr ptr;
+        NvCore::NvMap::Handle::Id handle;
+        DAddr ptr;
         u64 offset;
         u64 size;
         bool fixed;
         bool big_page; // Only valid if fixed == false
         bool sparse_alloc;
 
-        Mapping(VAddr ptr_, u64 offset_, u64 size_, bool fixed_, bool big_page_, bool sparse_alloc_)
+        Mapping(NvCore::NvMap::Handle::Id handle_, DAddr ptr_, u64 offset_, u64 size_, bool fixed_,
-            : ptr(ptr_), offset(offset_), size(size_), fixed(fixed_), big_page(big_page_),
+                bool big_page_, bool sparse_alloc_)
-              sparse_alloc(sparse_alloc_) {}
+            : handle(handle_), ptr(ptr_), offset(offset_), size(size_), fixed(fixed_),
+              big_page(big_page_), sparse_alloc(sparse_alloc_) {}
     };
 
     struct Allocation {
@@ -212,9 +214,7 @@ private:
         bool initialised{};
     } vm;
     std::shared_ptr<Tegra::MemoryManager> gmmu;
-
+    std::unordered_map<DeviceFD, size_t> sessions;
-    // s32 channel{};
-    // u32 big_page_size{VM::DEFAULT_BIG_PAGE_SIZE};
 };
 
 } // namespace Service::Nvidia::Devices

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

@@ -95,6 +95,9 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
     offset += SliceVectors(data, fence_thresholds, params.fence_count, offset);
 
     auto& gpu = system.GPU();
+    //auto& device_memory = system.Host1x().MemoryManager();
+    auto* session = core.GetSession(sessions[fd]);
+
     if (gpu.UseNvdec()) {
         for (std::size_t i = 0; i < syncpt_increments.size(); i++) {
             const SyncptIncr& syncpt_incr = syncpt_increments[i];
@@ -106,7 +109,7 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
         const auto object = nvmap.GetHandle(cmd_buffer.memory_id);
         ASSERT_OR_EXECUTE(object, return NvResult::InvalidState;);
         Tegra::ChCommandHeaderList cmdlist(cmd_buffer.word_count);
-        system.ApplicationMemory().ReadBlock(object->address + cmd_buffer.offset, cmdlist.data(),
+        session->process->GetMemory().ReadBlock(object->address + cmd_buffer.offset, cmdlist.data(),
                                              cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(core.Host1xDeviceFile().fd_to_id[fd], cmdlist);
     }
@@ -136,7 +139,8 @@ NvResult nvhost_nvdec_common::GetWaitbase(IoctlGetWaitbase& params) {
 NvResult nvhost_nvdec_common::MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries, DeviceFD fd) {
     const size_t num_entries = std::min(params.num_entries, static_cast<u32>(entries.size()));
     for (size_t i = 0; i < num_entries; i++) {
-        entries[i].map_address = nvmap.PinHandle(entries[i].map_handle, sessions[fd]);
+        DAddr pin_address = nvmap.PinHandle(entries[i].map_handle, sessions[fd], true);
+        entries[i].map_address = static_cast<u32>(pin_address);
     }
 
     return NvResult::Success;

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

@@ -123,8 +123,8 @@ NvResult nvmap::IocAlloc(IocAllocParams& params, DeviceFD fd) {
         return NvResult::InsufficientMemory;
     }
 
-    const auto result = file.AllocateHandle(params.handle, params.flags, params.align, params.kind,
+    const auto result =
-                                            params.address, sessions[fd]);
+        handle_description->Alloc(params.flags, params.align, params.kind, params.address);
     if (result != NvResult::Success) {
         LOG_CRITICAL(Service_NVDRV, "Object failed to allocate, handle={:08X}", params.handle);
         return result;

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

@@ -13,8 +13,6 @@
 #include "core/hle/service/nvdrv/nvdrv.h"
 #include "core/hle/service/nvdrv/nvdrv_interface.h"
 
-#pragma optimize("", off)
-
 namespace Service::Nvidia {
 
 void NVDRV::Open(HLERequestContext& ctx) {
@@ -173,8 +171,8 @@ void NVDRV::Initialize(HLERequestContext& ctx) {
     [[maybe_unused]] const auto transfer_memory_size = rp.Pop<u32>();
 
     auto& container = nvdrv->GetContainer();
-    auto process = ctx.GetObjectFromHandle(process_handle);
+    auto process = ctx.GetObjectFromHandle<Kernel::KProcess>(process_handle);
-    session_id = container.OpenSession(process->DynamicCast<Kernel::KProcess*>());
+    session_id = container.OpenSession(process.GetPointerUnsafe());
 
     is_initialized = true;
 }

src/core/memory.cpp

@@ -24,6 +24,8 @@
 #include "core/hle/kernel/k_process.h"
 #include "core/memory.h"
 #include "video_core/gpu.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/rasterizer_download_area.h"
 
 namespace Core::Memory {
@@ -638,15 +640,16 @@ struct Memory::Impl {
                   base * YUZU_PAGESIZE, (base + size) * YUZU_PAGESIZE);
 
         // During boot, current_page_table might not be set yet, in which case we need not flush
-        if (system.IsPoweredOn()) {
+        /*if (system.IsPoweredOn()) {
             auto& gpu = system.GPU();
             for (u64 i = 0; i < size; i++) {
                 const auto page = base + i;
                 if (page_table.pointers[page].Type() == Common::PageType::RasterizerCachedMemory) {
+
                     gpu.FlushAndInvalidateRegion(page << YUZU_PAGEBITS, YUZU_PAGESIZE);
                 }
             }
-        }
+        }*/
 
         const auto end = base + size;
         ASSERT_MSG(end <= page_table.pointers.size(), "out of range mapping at {:016X}",
@@ -811,10 +814,15 @@ struct Memory::Impl {
         return true;
     }
 
-    void HandleRasterizerDownload(VAddr address, size_t size) {
+    void HandleRasterizerDownload(VAddr v_address, size_t size) {
+        const auto* p = GetPointerImpl(
+            v_address, []() {}, []() {});
+        auto& gpu_device_memory = system.Host1x().MemoryManager();
+        DAddr address =
+            gpu_device_memory.GetAddressFromPAddr(system.DeviceMemory().GetRawPhysicalAddr(p));
         const size_t core = system.GetCurrentHostThreadID();
         auto& current_area = rasterizer_read_areas[core];
-        const VAddr end_address = address + size;
+        const DAddr end_address = address + size;
         if (current_area.start_address <= address && end_address <= current_area.end_address)
             [[likely]] {
             return;
@@ -822,7 +830,10 @@ struct Memory::Impl {
         current_area = system.GPU().OnCPURead(address, size);
     }
 
-    void HandleRasterizerWrite(VAddr address, size_t size) {
+    void HandleRasterizerWrite(VAddr v_address, size_t size) {
+        const auto* p = GetPointerImpl(
+            v_address, []() {}, []() {});
+        PAddr address = system.DeviceMemory().GetRawPhysicalAddr(p);
         constexpr size_t sys_core = Core::Hardware::NUM_CPU_CORES - 1;
         const size_t core = std::min(system.GetCurrentHostThreadID(),
                                      sys_core); // any other calls threads go to syscore.
@@ -836,7 +847,7 @@ struct Memory::Impl {
             }
         });
         auto& current_area = rasterizer_write_areas[core];
-        VAddr subaddress = address >> YUZU_PAGEBITS;
+        PAddr subaddress = address >> YUZU_PAGEBITS;
         bool do_collection = current_area.last_address == subaddress;
         if (!do_collection) [[unlikely]] {
             do_collection = system.GPU().OnCPUWrite(address, size);
@@ -849,7 +860,7 @@ struct Memory::Impl {
     }
 
     struct GPUDirtyState {
-        VAddr last_address;
+        PAddr last_address;
     };
 
     void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) {

src/core/memory.h

@@ -498,209 +498,4 @@ private:
     std::unique_ptr<Impl> impl;
 };
 
-enum GuestMemoryFlags : u32 {
-    Read = 1 << 0,
-    Write = 1 << 1,
-    Safe = 1 << 2,
-    Cached = 1 << 3,
-
-    SafeRead = Read | Safe,
-    SafeWrite = Write | Safe,
-    SafeReadWrite = SafeRead | SafeWrite,
-    SafeReadCachedWrite = SafeReadWrite | Cached,
-
-    UnsafeRead = Read,
-    UnsafeWrite = Write,
-    UnsafeReadWrite = UnsafeRead | UnsafeWrite,
-    UnsafeReadCachedWrite = UnsafeReadWrite | Cached,
-};
-
-namespace {
-template <typename M, typename T, GuestMemoryFlags FLAGS>
-class GuestMemory {
-    using iterator = T*;
-    using const_iterator = const T*;
-    using value_type = T;
-    using element_type = T;
-    using iterator_category = std::contiguous_iterator_tag;
-
-public:
-    GuestMemory() = delete;
-    explicit GuestMemory(M& memory, u64 addr, std::size_t size,
-                         Common::ScratchBuffer<T>* backup = nullptr)
-        : m_memory{memory}, m_addr{addr}, m_size{size} {
-        static_assert(FLAGS & GuestMemoryFlags::Read || FLAGS & GuestMemoryFlags::Write);
-        if constexpr (FLAGS & GuestMemoryFlags::Read) {
-            Read(addr, size, backup);
-        }
-    }
-
-    ~GuestMemory() = default;
-
-    T* data() noexcept {
-        return m_data_span.data();
-    }
-
-    const T* data() const noexcept {
-        return m_data_span.data();
-    }
-
-    size_t size() const noexcept {
-        return m_size;
-    }
-
-    size_t size_bytes() const noexcept {
-        return this->size() * sizeof(T);
-    }
-
-    [[nodiscard]] T* begin() noexcept {
-        return this->data();
-    }
-
-    [[nodiscard]] const T* begin() const noexcept {
-        return this->data();
-    }
-
-    [[nodiscard]] T* end() noexcept {
-        return this->data() + this->size();
-    }
-
-    [[nodiscard]] const T* end() const noexcept {
-        return this->data() + this->size();
-    }
-
-    T& operator[](size_t index) noexcept {
-        return m_data_span[index];
-    }
-
-    const T& operator[](size_t index) const noexcept {
-        return m_data_span[index];
-    }
-
-    void SetAddressAndSize(u64 addr, std::size_t size) noexcept {
-        m_addr = addr;
-        m_size = size;
-        m_addr_changed = true;
-    }
-
-    std::span<T> Read(u64 addr, std::size_t size,
-                      Common::ScratchBuffer<T>* backup = nullptr) noexcept {
-        m_addr = addr;
-        m_size = size;
-        if (m_size == 0) {
-            m_is_data_copy = true;
-            return {};
-        }
-
-        if (this->TrySetSpan()) {
-            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-                m_memory.FlushRegion(m_addr, this->size_bytes());
-            }
-        } else {
-            if (backup) {
-                backup->resize_destructive(this->size());
-                m_data_span = *backup;
-            } else {
-                m_data_copy.resize(this->size());
-                m_data_span = std::span(m_data_copy);
-            }
-            m_is_data_copy = true;
-            m_span_valid = true;
-            if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-                m_memory.ReadBlock(m_addr, this->data(), this->size_bytes());
-            } else {
-                m_memory.ReadBlockUnsafe(m_addr, this->data(), this->size_bytes());
-            }
-        }
-        return m_data_span;
-    }
-
-    void Write(std::span<T> write_data) noexcept {
-        if constexpr (FLAGS & GuestMemoryFlags::Cached) {
-            m_memory.WriteBlockCached(m_addr, write_data.data(), this->size_bytes());
-        } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-            m_memory.WriteBlock(m_addr, write_data.data(), this->size_bytes());
-        } else {
-            m_memory.WriteBlockUnsafe(m_addr, write_data.data(), this->size_bytes());
-        }
-    }
-
-    bool TrySetSpan() noexcept {
-        if (u8* ptr = m_memory.GetSpan(m_addr, this->size_bytes()); ptr) {
-            m_data_span = {reinterpret_cast<T*>(ptr), this->size()};
-            m_span_valid = true;
-            return true;
-        }
-        return false;
-    }
-
-protected:
-    bool IsDataCopy() const noexcept {
-        return m_is_data_copy;
-    }
-
-    bool AddressChanged() const noexcept {
-        return m_addr_changed;
-    }
-
-    M& m_memory;
-    u64 m_addr{};
-    size_t m_size{};
-    std::span<T> m_data_span{};
-    std::vector<T> m_data_copy{};
-    bool m_span_valid{false};
-    bool m_is_data_copy{false};
-    bool m_addr_changed{false};
-};
-
-template <typename M, typename T, GuestMemoryFlags FLAGS>
-class GuestMemoryScoped : public GuestMemory<M, T, FLAGS> {
-public:
-    GuestMemoryScoped() = delete;
-    explicit GuestMemoryScoped(M& memory, u64 addr, std::size_t size,
-                               Common::ScratchBuffer<T>* backup = nullptr)
-        : GuestMemory<M, T, FLAGS>(memory, addr, size, backup) {
-        if constexpr (!(FLAGS & GuestMemoryFlags::Read)) {
-            if (!this->TrySetSpan()) {
-                if (backup) {
-                    this->m_data_span = *backup;
-                    this->m_span_valid = true;
-                    this->m_is_data_copy = true;
-                }
-            }
-        }
-    }
-
-    ~GuestMemoryScoped() {
-        if constexpr (FLAGS & GuestMemoryFlags::Write) {
-            if (this->size() == 0) [[unlikely]] {
-                return;
-            }
-
-            if (this->AddressChanged() || this->IsDataCopy()) {
-                ASSERT(this->m_span_valid);
-                if constexpr (FLAGS & GuestMemoryFlags::Cached) {
-                    this->m_memory.WriteBlockCached(this->m_addr, this->data(), this->size_bytes());
-                } else if constexpr (FLAGS & GuestMemoryFlags::Safe) {
-                    this->m_memory.WriteBlock(this->m_addr, this->data(), this->size_bytes());
-                } else {
-                    this->m_memory.WriteBlockUnsafe(this->m_addr, this->data(), this->size_bytes());
-                }
-            } else if constexpr ((FLAGS & GuestMemoryFlags::Safe) ||
-                                 (FLAGS & GuestMemoryFlags::Cached)) {
-                this->m_memory.InvalidateRegion(this->m_addr, this->size_bytes());
-            }
-        }
-    }
-};
-} // namespace
-
-template <typename T, GuestMemoryFlags FLAGS>
-using CpuGuestMemory = GuestMemory<Memory, T, FLAGS>;
-template <typename T, GuestMemoryFlags FLAGS>
-using CpuGuestMemoryScoped = GuestMemoryScoped<Memory, T, FLAGS>;
-template <typename T, GuestMemoryFlags FLAGS>
-using GpuGuestMemory = GuestMemory<Tegra::MemoryManager, T, FLAGS>;
-template <typename T, GuestMemoryFlags FLAGS>
-using GpuGuestMemoryScoped = GuestMemoryScoped<Tegra::MemoryManager, T, FLAGS>;
 } // namespace Core::Memory

src/video_core/CMakeLists.txt

@@ -95,6 +95,7 @@ add_library(video_core STATIC
     gpu.h
     gpu_thread.cpp
     gpu_thread.h
+    guest_memory.h
     invalidation_accumulator.h
     memory_manager.cpp
     memory_manager.h
@@ -107,8 +108,6 @@ add_library(video_core STATIC
     query_cache/query_stream.h
     query_cache/types.h
     query_cache.h
-    rasterizer_accelerated.cpp
-    rasterizer_accelerated.h
     rasterizer_interface.h
     renderer_base.cpp
     renderer_base.h

src/video_core/buffer_cache/buffer_base.h

@@ -33,13 +33,12 @@ struct NullBufferParams {};
  *
  * The buffer size and address is forcefully aligned to CPU page boundaries.
  */
-template <class RasterizerInterface>
 class BufferBase {
 public:
     static constexpr u64 BASE_PAGE_BITS = 16;
     static constexpr u64 BASE_PAGE_SIZE = 1ULL << BASE_PAGE_BITS;
 
-    explicit BufferBase(RasterizerInterface& rasterizer_, VAddr cpu_addr_, u64 size_bytes_)
+    explicit BufferBase(VAddr cpu_addr_, u64 size_bytes_)
         : cpu_addr{cpu_addr_}, size_bytes{size_bytes_} {}
 
     explicit BufferBase(NullBufferParams) {}

src/video_core/buffer_cache/buffer_cache_base.h

@@ -32,7 +32,6 @@
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
 #include "common/settings.h"
-#include "core/memory.h"
 #include "video_core/buffer_cache/buffer_base.h"
 #include "video_core/control/channel_state_cache.h"
 #include "video_core/delayed_destruction_ring.h"
@@ -41,7 +40,6 @@
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/memory_manager.h"
-#include "video_core/rasterizer_interface.h"
 #include "video_core/surface.h"
 #include "video_core/texture_cache/slot_vector.h"
 #include "video_core/texture_cache/types.h"
@@ -94,7 +92,7 @@ static constexpr BufferId NULL_BUFFER_ID{0};
 static constexpr u32 DEFAULT_SKIP_CACHE_SIZE = static_cast<u32>(4_KiB);
 
 struct Binding {
-    VAddr cpu_addr{};
+    DAddr device_addr{};
     u32 size{};
     BufferId buffer_id;
 };
@@ -104,7 +102,7 @@ struct TextureBufferBinding : Binding {
 };
 
 static constexpr Binding NULL_BINDING{
-    .cpu_addr = 0,
+    .device_addr = 0,
     .size = 0,
     .buffer_id = NULL_BUFFER_ID,
 };
@@ -204,10 +202,10 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<BufferCacheChannelInf
     using Async_Buffer = typename P::Async_Buffer;
     using MemoryTracker = typename P::MemoryTracker;
 
-    using IntervalCompare = std::less<VAddr>;
+    using IntervalCompare = std::less<DAddr>;
-    using IntervalInstance = boost::icl::interval_type_default<VAddr, std::less>;
+    using IntervalInstance = boost::icl::interval_type_default<DAddr, std::less>;
-    using IntervalAllocator = boost::fast_pool_allocator<VAddr>;
+    using IntervalAllocator = boost::fast_pool_allocator<DAddr>;
-    using IntervalSet = boost::icl::interval_set<VAddr>;
+    using IntervalSet = boost::icl::interval_set<DAddr>;
     using IntervalType = typename IntervalSet::interval_type;
 
     template <typename Type>
@@ -230,32 +228,31 @@ class BufferCache : public VideoCommon::ChannelSetupCaches<BufferCacheChannelInf
 
     using OverlapCombine = counter_add_functor<int>;
     using OverlapSection = boost::icl::inter_section<int>;
-    using OverlapCounter = boost::icl::split_interval_map<VAddr, int>;
+    using OverlapCounter = boost::icl::split_interval_map<DAddr, int>;
 
     struct OverlapResult {
         boost::container::small_vector<BufferId, 16> ids;
-        VAddr begin;
+        DAddr begin;
-        VAddr end;
+        DAddr end;
         bool has_stream_leap = false;
     };
 
 public:
-    explicit BufferCache(VideoCore::RasterizerInterface& rasterizer_,
+    explicit BufferCache(Tegra::MaxwellDeviceMemoryManager& device_memory_, Runtime& runtime_);
-                         Core::Memory::Memory& cpu_memory_, Runtime& runtime_);
 
     void TickFrame();
 
-    void WriteMemory(VAddr cpu_addr, u64 size);
+    void WriteMemory(DAddr device_addr, u64 size);
 
-    void CachedWriteMemory(VAddr cpu_addr, u64 size);
+    void CachedWriteMemory(DAddr device_addr, u64 size);
 
-    bool OnCPUWrite(VAddr cpu_addr, u64 size);
+    bool OnCPUWrite(DAddr device_addr, u64 size);
 
-    void DownloadMemory(VAddr cpu_addr, u64 size);
+    void DownloadMemory(DAddr device_addr, u64 size);
 
-    std::optional<VideoCore::RasterizerDownloadArea> GetFlushArea(VAddr cpu_addr, u64 size);
+    std::optional<VideoCore::RasterizerDownloadArea> GetFlushArea(DAddr device_addr, u64 size);
 
-    bool InlineMemory(VAddr dest_address, size_t copy_size, std::span<const u8> inlined_buffer);
+    bool InlineMemory(DAddr dest_address, size_t copy_size, std::span<const u8> inlined_buffer);
 
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size);
 
@@ -300,7 +297,7 @@ public:
                                                        ObtainBufferSynchronize sync_info,
                                                        ObtainBufferOperation post_op);
 
-    [[nodiscard]] std::pair<Buffer*, u32> ObtainCPUBuffer(VAddr gpu_addr, u32 size,
+    [[nodiscard]] std::pair<Buffer*, u32> ObtainCPUBuffer(DAddr gpu_addr, u32 size,
                                                           ObtainBufferSynchronize sync_info,
                                                           ObtainBufferOperation post_op);
     void FlushCachedWrites();
@@ -326,13 +323,13 @@ public:
     bool DMAClear(GPUVAddr src_address, u64 amount, u32 value);
 
     /// Return true when a CPU region is modified from the GPU
-    [[nodiscard]] bool IsRegionGpuModified(VAddr addr, size_t size);
+    [[nodiscard]] bool IsRegionGpuModified(DAddr addr, size_t size);
 
     /// Return true when a region is registered on the cache
-    [[nodiscard]] bool IsRegionRegistered(VAddr addr, size_t size);
+    [[nodiscard]] bool IsRegionRegistered(DAddr addr, size_t size);
 
     /// Return true when a CPU region is modified from the CPU
-    [[nodiscard]] bool IsRegionCpuModified(VAddr addr, size_t size);
+    [[nodiscard]] bool IsRegionCpuModified(DAddr addr, size_t size);
 
     void SetDrawIndirect(
         const Tegra::Engines::DrawManager::IndirectParams* current_draw_indirect_) {
@@ -366,9 +363,9 @@ private:
     }
 
     template <typename Func>
-    void ForEachBufferInRange(VAddr cpu_addr, u64 size, Func&& func) {
+    void ForEachBufferInRange(DAddr device_addr, u64 size, Func&& func) {
-        const u64 page_end = Common::DivCeil(cpu_addr + size, CACHING_PAGESIZE);
+        const u64 page_end = Common::DivCeil(device_addr + size, CACHING_PAGESIZE);
-        for (u64 page = cpu_addr >> CACHING_PAGEBITS; page < page_end;) {
+        for (u64 page = device_addr >> CACHING_PAGEBITS; page < page_end;) {
             const BufferId buffer_id = page_table[page];
             if (!buffer_id) {
                 ++page;
@@ -377,15 +374,15 @@ private:
             Buffer& buffer = slot_buffers[buffer_id];
             func(buffer_id, buffer);
 
-            const VAddr end_addr = buffer.CpuAddr() + buffer.SizeBytes();
+            const DAddr end_addr = buffer.CpuAddr() + buffer.SizeBytes();
             page = Common::DivCeil(end_addr, CACHING_PAGESIZE);
         }
     }
 
     template <typename Func>
-    void ForEachInRangeSet(IntervalSet& current_range, VAddr cpu_addr, u64 size, Func&& func) {
+    void ForEachInRangeSet(IntervalSet& current_range, DAddr device_addr, u64 size, Func&& func) {
-        const VAddr start_address = cpu_addr;
+        const DAddr start_address = device_addr;
-        const VAddr end_address = start_address + size;
+        const DAddr end_address = start_address + size;
         const IntervalType search_interval{start_address, end_address};
         auto it = current_range.lower_bound(search_interval);
         if (it == current_range.end()) {
@@ -393,8 +390,8 @@ private:
         }
         auto end_it = current_range.upper_bound(search_interval);
         for (; it != end_it; it++) {
-            VAddr inter_addr_end = it->upper();
+            DAddr inter_addr_end = it->upper();
-            VAddr inter_addr = it->lower();
+            DAddr inter_addr = it->lower();
             if (inter_addr_end > end_address) {
                 inter_addr_end = end_address;
             }
@@ -406,10 +403,10 @@ private:
     }
 
     template <typename Func>
-    void ForEachInOverlapCounter(OverlapCounter& current_range, VAddr cpu_addr, u64 size,
+    void ForEachInOverlapCounter(OverlapCounter& current_range, DAddr device_addr, u64 size,
                                  Func&& func) {
-        const VAddr start_address = cpu_addr;
+        const DAddr start_address = device_addr;
-        const VAddr end_address = start_address + size;
+        const DAddr end_address = start_address + size;
         const IntervalType search_interval{start_address, end_address};
         auto it = current_range.lower_bound(search_interval);
         if (it == current_range.end()) {
@@ -418,8 +415,8 @@ private:
         auto end_it = current_range.upper_bound(search_interval);
         for (; it != end_it; it++) {
             auto& inter = it->first;
-            VAddr inter_addr_end = inter.upper();
+            DAddr inter_addr_end = inter.upper();
-            VAddr inter_addr = inter.lower();
+            DAddr inter_addr = inter.lower();
             if (inter_addr_end > end_address) {
                 inter_addr_end = end_address;
             }
@@ -451,9 +448,9 @@ private:
         } while (any_removals);
     }
 
-    static bool IsRangeGranular(VAddr cpu_addr, size_t size) {
+    static bool IsRangeGranular(DAddr device_addr, size_t size) {
-        return (cpu_addr & ~Core::Memory::YUZU_PAGEMASK) ==
+        return (device_addr & ~Core::Memory::YUZU_PAGEMASK) ==
-               ((cpu_addr + size) & ~Core::Memory::YUZU_PAGEMASK);
+               ((device_addr + size) & ~Core::Memory::YUZU_PAGEMASK);
     }
 
     void RunGarbageCollector();
@@ -508,15 +505,15 @@ private:
 
     void UpdateComputeTextureBuffers();
 
-    void MarkWrittenBuffer(BufferId buffer_id, VAddr cpu_addr, u32 size);
+    void MarkWrittenBuffer(BufferId buffer_id, DAddr device_addr, u32 size);
 
-    [[nodiscard]] BufferId FindBuffer(VAddr cpu_addr, u32 size);
+    [[nodiscard]] BufferId FindBuffer(DAddr device_addr, u32 size);
 
-    [[nodiscard]] OverlapResult ResolveOverlaps(VAddr cpu_addr, u32 wanted_size);
+    [[nodiscard]] OverlapResult ResolveOverlaps(DAddr device_addr, u32 wanted_size);
 
     void JoinOverlap(BufferId new_buffer_id, BufferId overlap_id, bool accumulate_stream_score);
 
-    [[nodiscard]] BufferId CreateBuffer(VAddr cpu_addr, u32 wanted_size);
+    [[nodiscard]] BufferId CreateBuffer(DAddr device_addr, u32 wanted_size);
 
     void Register(BufferId buffer_id);
 
@@ -527,7 +524,7 @@ private:
 
     void TouchBuffer(Buffer& buffer, BufferId buffer_id) noexcept;
 
-    bool SynchronizeBuffer(Buffer& buffer, VAddr cpu_addr, u32 size);
+    bool SynchronizeBuffer(Buffer& buffer, DAddr device_addr, u32 size);
 
     void UploadMemory(Buffer& buffer, u64 total_size_bytes, u64 largest_copy,
                       std::span<BufferCopy> copies);
@@ -539,7 +536,7 @@ private:
 
     void DownloadBufferMemory(Buffer& buffer_id);
 
-    void DownloadBufferMemory(Buffer& buffer_id, VAddr cpu_addr, u64 size);
+    void DownloadBufferMemory(Buffer& buffer_id, DAddr device_addr, u64 size);
 
     void DeleteBuffer(BufferId buffer_id, bool do_not_mark = false);
 
@@ -549,7 +546,7 @@ private:
     [[nodiscard]] TextureBufferBinding GetTextureBufferBinding(GPUVAddr gpu_addr, u32 size,
                                                                PixelFormat format);
 
-    [[nodiscard]] std::span<const u8> ImmediateBufferWithData(VAddr cpu_addr, size_t size);
+    [[nodiscard]] std::span<const u8> ImmediateBufferWithData(DAddr device_addr, size_t size);
 
     [[nodiscard]] std::span<u8> ImmediateBuffer(size_t wanted_capacity);
 
@@ -557,11 +554,10 @@ private:
 
     void ClearDownload(IntervalType subtract_interval);
 
-    void InlineMemoryImplementation(VAddr dest_address, size_t copy_size,
+    void InlineMemoryImplementation(DAddr dest_address, size_t copy_size,
                                     std::span<const u8> inlined_buffer);
 
-    VideoCore::RasterizerInterface& rasterizer;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
-    Core::Memory::Memory& cpu_memory;
 
     SlotVector<Buffer> slot_buffers;
     DelayedDestructionRing<Buffer, 8> delayed_destruction_ring;
@@ -598,7 +594,7 @@ private:
     u64 critical_memory = 0;
     BufferId inline_buffer_id;
 
-    std::array<BufferId, ((1ULL << 39) >> CACHING_PAGEBITS)> page_table;
+    std::array<BufferId, ((1ULL << 34) >> CACHING_PAGEBITS)> page_table;
     Common::ScratchBuffer<u8> tmp_buffer;
 };
 

src/video_core/buffer_cache/memory_tracker_base.h

@@ -17,19 +17,19 @@
 
 namespace VideoCommon {
 
-template <class RasterizerInterface>
+template <typename DeviceTracker>
 class MemoryTrackerBase {
-    static constexpr size_t MAX_CPU_PAGE_BITS = 39;
+    static constexpr size_t MAX_CPU_PAGE_BITS = 34;
     static constexpr size_t HIGHER_PAGE_BITS = 22;
     static constexpr size_t HIGHER_PAGE_SIZE = 1ULL << HIGHER_PAGE_BITS;
     static constexpr size_t HIGHER_PAGE_MASK = HIGHER_PAGE_SIZE - 1ULL;
     static constexpr size_t NUM_HIGH_PAGES = 1ULL << (MAX_CPU_PAGE_BITS - HIGHER_PAGE_BITS);
     static constexpr size_t MANAGER_POOL_SIZE = 32;
     static constexpr size_t WORDS_STACK_NEEDED = HIGHER_PAGE_SIZE / BYTES_PER_WORD;
-    using Manager = WordManager<RasterizerInterface, WORDS_STACK_NEEDED>;
+    using Manager = WordManager<DeviceTracker, WORDS_STACK_NEEDED>;
 
 public:
-    MemoryTrackerBase(RasterizerInterface& rasterizer_) : rasterizer{&rasterizer_} {}
+    MemoryTrackerBase(DeviceTracker& device_tracker_) : device_tracker{&device_tracker_} {}
     ~MemoryTrackerBase() = default;
 
     /// Returns the inclusive CPU modified range in a begin end pair
@@ -74,7 +74,7 @@ public:
             });
     }
 
-    /// Mark region as CPU modified, notifying the rasterizer about this change
+    /// Mark region as CPU modified, notifying the device_tracker about this change
     void MarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 query_size) {
         IteratePages<true>(dirty_cpu_addr, query_size,
                            [](Manager* manager, u64 offset, size_t size) {
@@ -83,7 +83,7 @@ public:
                            });
     }
 
-    /// Unmark region as CPU modified, notifying the rasterizer about this change
+    /// Unmark region as CPU modified, notifying the device_tracker about this change
     void UnmarkRegionAsCpuModified(VAddr dirty_cpu_addr, u64 query_size) {
         IteratePages<true>(dirty_cpu_addr, query_size,
                            [](Manager* manager, u64 offset, size_t size) {
@@ -139,7 +139,7 @@ public:
             });
     }
 
-    /// Flushes cached CPU writes, and notify the rasterizer about the deltas
+    /// Flushes cached CPU writes, and notify the device_tracker about the deltas
     void FlushCachedWrites(VAddr query_cpu_addr, u64 query_size) noexcept {
         IteratePages<false>(query_cpu_addr, query_size,
                             [](Manager* manager, [[maybe_unused]] u64 offset,
@@ -280,7 +280,7 @@ private:
         manager_pool.emplace_back();
         auto& last_pool = manager_pool.back();
         for (size_t i = 0; i < MANAGER_POOL_SIZE; i++) {
-            new (&last_pool[i]) Manager(0, *rasterizer, HIGHER_PAGE_SIZE);
+            new (&last_pool[i]) Manager(0, *device_tracker, HIGHER_PAGE_SIZE);
             free_managers.push_back(&last_pool[i]);
         }
         return on_return();
@@ -293,7 +293,7 @@ private:
 
     std::unordered_set<u32> cached_pages;
 
-    RasterizerInterface* rasterizer = nullptr;
+    DeviceTracker* device_tracker = nullptr;
 };
 
 } // namespace VideoCommon

src/video_core/buffer_cache/word_manager.h

@@ -163,11 +163,11 @@ struct Words {
     WordsArray<stack_words> preflushable;
 };
 
-template <class RasterizerInterface, size_t stack_words = 1>
+template <class DeviceTracker, size_t stack_words = 1>
 class WordManager {
 public:
-    explicit WordManager(VAddr cpu_addr_, RasterizerInterface& rasterizer_, u64 size_bytes)
+    explicit WordManager(VAddr cpu_addr_, DeviceTracker& tracker_, u64 size_bytes)
-        : cpu_addr{cpu_addr_}, rasterizer{&rasterizer_}, words{size_bytes} {}
+        : cpu_addr{cpu_addr_}, tracker{&tracker_}, words{size_bytes} {}
 
     explicit WordManager() = default;
 
@@ -279,7 +279,7 @@ public:
     }
 
     /**
-     * Loop over each page in the given range, turn off those bits and notify the rasterizer if
+     * Loop over each page in the given range, turn off those bits and notify the tracker if
      * needed. Call the given function on each turned off range.
      *
      * @param query_cpu_range Base CPU address to loop over
@@ -459,26 +459,26 @@ private:
     }
 
     /**
-     * Notify rasterizer about changes in the CPU tracking state of a word in the buffer
+     * Notify tracker about changes in the CPU tracking state of a word in the buffer
      *
-     * @param word_index   Index to the word to notify to the rasterizer
+     * @param word_index   Index to the word to notify to the tracker
      * @param current_bits Current state of the word
      * @param new_bits     New state of the word
      *
-     * @tparam add_to_rasterizer True when the rasterizer should start tracking the new pages
+     * @tparam add_to_tracker True when the tracker should start tracking the new pages
      */
-    template <bool add_to_rasterizer>
+    template <bool add_to_tracker>
     void NotifyRasterizer(u64 word_index, u64 current_bits, u64 new_bits) const {
-        u64 changed_bits = (add_to_rasterizer ? current_bits : ~current_bits) & new_bits;
+        u64 changed_bits = (add_to_tracker ? current_bits : ~current_bits) & new_bits;
         VAddr addr = cpu_addr + word_index * BYTES_PER_WORD;
         IteratePages(changed_bits, [&](size_t offset, size_t size) {
-            rasterizer->UpdatePagesCachedCount(addr + offset * BYTES_PER_PAGE,
+            tracker->UpdatePagesCachedCount(addr + offset * BYTES_PER_PAGE,
-                                               size * BYTES_PER_PAGE, add_to_rasterizer ? 1 : -1);
+                                               size * BYTES_PER_PAGE, add_to_tracker ? 1 : -1);
         });
     }
 
     VAddr cpu_addr = 0;
-    RasterizerInterface* rasterizer = nullptr;
+    DeviceTracker* tracker = nullptr;
     Words<stack_words> words;
 };
 

src/video_core/dma_pusher.cpp

@@ -5,10 +5,10 @@
 #include "common/microprofile.h"
 #include "common/settings.h"
 #include "core/core.h"
-#include "core/memory.h"
 #include "video_core/dma_pusher.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
+#include "video_core/guest_memory.h"
 #include "video_core/memory_manager.h"
 
 namespace Tegra {
@@ -85,15 +85,15 @@ bool DmaPusher::Step() {
             }
         }
         const auto safe_process = [&] {
-            Core::Memory::GpuGuestMemory<Tegra::CommandHeader,
+            Tegra::Memory::GpuGuestMemory<Tegra::CommandHeader,
-                                         Core::Memory::GuestMemoryFlags::SafeRead>
+                                         Tegra::Memory::GuestMemoryFlags::SafeRead>
                 headers(memory_manager, dma_state.dma_get, command_list_header.size,
                         &command_headers);
             ProcessCommands(headers);
         };
         const auto unsafe_process = [&] {
-            Core::Memory::GpuGuestMemory<Tegra::CommandHeader,
+            Tegra::Memory::GpuGuestMemory<Tegra::CommandHeader,
-                                         Core::Memory::GuestMemoryFlags::UnsafeRead>
+                                         Tegra::Memory::GuestMemoryFlags::UnsafeRead>
                 headers(memory_manager, dma_state.dma_get, command_list_header.size,
                         &command_headers);
             ProcessCommands(headers);

src/video_core/engines/engine_upload.cpp

@@ -5,8 +5,8 @@
 
 #include "common/algorithm.h"
 #include "common/assert.h"
-#include "core/memory.h"
 #include "video_core/engines/engine_upload.h"
+#include "video_core/guest_memory.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/textures/decoders.h"
@@ -68,7 +68,8 @@ void State::ProcessData(std::span<const u8> read_buffer) {
             true, bytes_per_pixel, width, regs.dest.height, regs.dest.depth,
             regs.dest.BlockHeight(), regs.dest.BlockDepth());
 
-        Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
+        Tegra::Memory::GpuGuestMemoryScoped<u8,
+                                            Tegra::Memory::GuestMemoryFlags::SafeReadCachedWrite>
             tmp(memory_manager, address, dst_size, &tmp_buffer);
 
         Tegra::Texture::SwizzleSubrect(tmp, read_buffer, bytes_per_pixel, width, regs.dest.height,

src/video_core/engines/maxwell_dma.cpp

@@ -11,6 +11,7 @@
 #include "core/memory.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/maxwell_dma.h"
+#include "video_core/guest_memory.h"
 #include "video_core/memory_manager.h"
 #include "video_core/renderer_base.h"
 #include "video_core/textures/decoders.h"
@@ -133,8 +134,8 @@ void MaxwellDMA::Launch() {
                 UNIMPLEMENTED_IF(regs.offset_out % 16 != 0);
                 read_buffer.resize_destructive(16);
                 for (u32 offset = 0; offset < regs.line_length_in; offset += 16) {
-                    Core::Memory::GpuGuestMemoryScoped<
+                    Tegra::Memory::GpuGuestMemoryScoped<
-                        u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
+                        u8, Tegra::Memory::GuestMemoryFlags::SafeReadCachedWrite>
                         tmp_write_buffer(memory_manager,
                                          convert_linear_2_blocklinear_addr(regs.offset_in + offset),
                                          16, &read_buffer);
@@ -146,16 +147,16 @@ void MaxwellDMA::Launch() {
                 UNIMPLEMENTED_IF(regs.offset_out % 16 != 0);
                 read_buffer.resize_destructive(16);
                 for (u32 offset = 0; offset < regs.line_length_in; offset += 16) {
-                    Core::Memory::GpuGuestMemoryScoped<
+                    Tegra::Memory::GpuGuestMemoryScoped<
-                        u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
+                        u8, Tegra::Memory::GuestMemoryFlags::SafeReadCachedWrite>
                         tmp_write_buffer(memory_manager, regs.offset_in + offset, 16, &read_buffer);
                     tmp_write_buffer.SetAddressAndSize(
                         convert_linear_2_blocklinear_addr(regs.offset_out + offset), 16);
                 }
             } else {
                 if (!accelerate.BufferCopy(regs.offset_in, regs.offset_out, regs.line_length_in)) {
-                    Core::Memory::GpuGuestMemoryScoped<
+                    Tegra::Memory::GpuGuestMemoryScoped<
-                        u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
+                        u8, Tegra::Memory::GuestMemoryFlags::SafeReadCachedWrite>
                         tmp_write_buffer(memory_manager, regs.offset_in, regs.line_length_in,
                                          &read_buffer);
                     tmp_write_buffer.SetAddressAndSize(regs.offset_out, regs.line_length_in);
@@ -226,9 +227,9 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
 
     const size_t dst_size = dst_operand.pitch * regs.line_count;
 
-    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
+    Tegra::Memory::GpuGuestMemory<u8, Tegra::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
         memory_manager, src_operand.address, src_size, &read_buffer);
-    Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::UnsafeReadCachedWrite>
+    Tegra::Memory::GpuGuestMemoryScoped<u8, Tegra::Memory::GuestMemoryFlags::UnsafeReadCachedWrite>
         tmp_write_buffer(memory_manager, dst_operand.address, dst_size, &write_buffer);
 
     UnswizzleSubrect(tmp_write_buffer, tmp_read_buffer, bytes_per_pixel, width, height, depth,
@@ -290,9 +291,9 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
 
     GPUVAddr src_addr = regs.offset_in;
     GPUVAddr dst_addr = regs.offset_out;
-    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
+    Tegra::Memory::GpuGuestMemory<u8, Tegra::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
         memory_manager, src_addr, src_size, &read_buffer);
-    Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::UnsafeReadCachedWrite>
+    Tegra::Memory::GpuGuestMemoryScoped<u8, Tegra::Memory::GuestMemoryFlags::UnsafeReadCachedWrite>
         tmp_write_buffer(memory_manager, dst_addr, dst_size, &write_buffer);
 
     //  If the input is linear and the output is tiled, swizzle the input and copy it over.
@@ -344,9 +345,9 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
 
     intermediate_buffer.resize_destructive(mid_buffer_size);
 
-    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
+    Tegra::Memory::GpuGuestMemory<u8, Tegra::Memory::GuestMemoryFlags::SafeRead> tmp_read_buffer(
         memory_manager, regs.offset_in, src_size, &read_buffer);
-    Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadCachedWrite>
+    Tegra::Memory::GpuGuestMemoryScoped<u8, Tegra::Memory::GuestMemoryFlags::SafeReadCachedWrite>
         tmp_write_buffer(memory_manager, regs.offset_out, dst_size, &write_buffer);
 
     UnswizzleSubrect(intermediate_buffer, tmp_read_buffer, bytes_per_pixel, src_width, src.height,

src/video_core/engines/sw_blitter/blitter.cpp

@@ -11,6 +11,7 @@
 #include "video_core/memory_manager.h"
 #include "video_core/surface.h"
 #include "video_core/textures/decoders.h"
+#include "video_core/guest_memory.h"
 
 namespace Tegra {
 class MemoryManager;
@@ -160,7 +161,7 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
     const auto dst_bytes_per_pixel = BytesPerBlock(PixelFormatFromRenderTargetFormat(dst.format));
     const size_t src_size = get_surface_size(src, src_bytes_per_pixel);
 
-    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::SafeRead> tmp_buffer(
+    Tegra::Memory::GpuGuestMemory<u8, Tegra::Memory::GuestMemoryFlags::SafeRead> tmp_buffer(
         memory_manager, src.Address(), src_size, &impl->tmp_buffer);
 
     const size_t src_copy_size = src_extent_x * src_extent_y * src_bytes_per_pixel;
@@ -220,7 +221,7 @@ bool SoftwareBlitEngine::Blit(Fermi2D::Surface& src, Fermi2D::Surface& dst,
     }
 
     const size_t dst_size = get_surface_size(dst, dst_bytes_per_pixel);
-    Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::SafeReadWrite>
+    Tegra::Memory::GpuGuestMemoryScoped<u8, Tegra::Memory::GuestMemoryFlags::SafeReadWrite>
         tmp_buffer2(memory_manager, dst.Address(), dst_size, &impl->tmp_buffer);
 
     if (dst.linear == Fermi2D::MemoryLayout::BlockLinear) {

src/video_core/framebuffer_config.h

@@ -14,7 +14,7 @@ namespace Tegra {
  * Struct describing framebuffer configuration
  */
 struct FramebufferConfig {
-    VAddr address{};
+    DAddr address{};
     u32 offset{};
     u32 width{};
     u32 height{};

src/video_core/gpu.cpp

@@ -34,6 +34,8 @@
 #include "video_core/renderer_base.h"
 #include "video_core/shader_notify.h"
 
+#pragma optimize("", off)
+
 namespace Tegra {
 
 struct GPU::Impl {
@@ -95,8 +97,8 @@ struct GPU::Impl {
 
     /// Synchronizes CPU writes with Host GPU memory.
     void InvalidateGPUCache() {
-        std::function<void(VAddr, size_t)> callback_writes(
+        std::function<void(PAddr, size_t)> callback_writes(
-            [this](VAddr address, size_t size) { rasterizer->OnCacheInvalidation(address, size); });
+            [this](PAddr address, size_t size) { rasterizer->OnCacheInvalidation(address, size); });
         system.GatherGPUDirtyMemory(callback_writes);
     }
 
@@ -279,11 +281,11 @@ struct GPU::Impl {
     }
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    void FlushRegion(VAddr addr, u64 size) {
+    void FlushRegion(DAddr addr, u64 size) {
         gpu_thread.FlushRegion(addr, size);
     }
 
-    VideoCore::RasterizerDownloadArea OnCPURead(VAddr addr, u64 size) {
+    VideoCore::RasterizerDownloadArea OnCPURead(DAddr addr, u64 size) {
         auto raster_area = rasterizer->GetFlushArea(addr, size);
         if (raster_area.preemtive) {
             return raster_area;
@@ -299,16 +301,16 @@ struct GPU::Impl {
     }
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    void InvalidateRegion(VAddr addr, u64 size) {
+    void InvalidateRegion(DAddr addr, u64 size) {
         gpu_thread.InvalidateRegion(addr, size);
     }
 
-    bool OnCPUWrite(VAddr addr, u64 size) {
+    bool OnCPUWrite(DAddr addr, u64 size) {
         return rasterizer->OnCPUWrite(addr, size);
     }
 
     /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
-    void FlushAndInvalidateRegion(VAddr addr, u64 size) {
+    void FlushAndInvalidateRegion(DAddr addr, u64 size) {
         gpu_thread.FlushAndInvalidateRegion(addr, size);
     }
 
@@ -437,7 +439,7 @@ void GPU::OnCommandListEnd() {
     impl->OnCommandListEnd();
 }
 
-u64 GPU::RequestFlush(VAddr addr, std::size_t size) {
+u64 GPU::RequestFlush(DAddr addr, std::size_t size) {
     return impl->RequestSyncOperation(
         [this, addr, size]() { impl->rasterizer->FlushRegion(addr, size); });
 }
@@ -557,23 +559,23 @@ void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     impl->SwapBuffers(framebuffer);
 }
 
-VideoCore::RasterizerDownloadArea GPU::OnCPURead(VAddr addr, u64 size) {
+VideoCore::RasterizerDownloadArea GPU::OnCPURead(PAddr addr, u64 size) {
     return impl->OnCPURead(addr, size);
 }
 
-void GPU::FlushRegion(VAddr addr, u64 size) {
+void GPU::FlushRegion(DAddr addr, u64 size) {
     impl->FlushRegion(addr, size);
 }
 
-void GPU::InvalidateRegion(VAddr addr, u64 size) {
+void GPU::InvalidateRegion(DAddr addr, u64 size) {
     impl->InvalidateRegion(addr, size);
 }
 
-bool GPU::OnCPUWrite(VAddr addr, u64 size) {
+bool GPU::OnCPUWrite(DAddr addr, u64 size) {
     return impl->OnCPUWrite(addr, size);
 }
 
-void GPU::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+void GPU::FlushAndInvalidateRegion(DAddr addr, u64 size) {
     impl->FlushAndInvalidateRegion(addr, size);
 }
 

src/video_core/gpu.h

@@ -158,7 +158,7 @@ public:
     void InitAddressSpace(Tegra::MemoryManager& memory_manager);
 
     /// Request a host GPU memory flush from the CPU.
-    [[nodiscard]] u64 RequestFlush(VAddr addr, std::size_t size);
+    [[nodiscard]] u64 RequestFlush(DAddr addr, std::size_t size);
 
     /// Obtains current flush request fence id.
     [[nodiscard]] u64 CurrentSyncRequestFence() const;
@@ -242,20 +242,20 @@ public:
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    [[nodiscard]] VideoCore::RasterizerDownloadArea OnCPURead(VAddr addr, u64 size);
+    [[nodiscard]] VideoCore::RasterizerDownloadArea OnCPURead(DAddr addr, u64 size);
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    void FlushRegion(VAddr addr, u64 size);
+    void FlushRegion(DAddr addr, u64 size);
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    void InvalidateRegion(VAddr addr, u64 size);
+    void InvalidateRegion(DAddr addr, u64 size);
 
     /// Notify rasterizer that CPU is trying to write this area. It returns true if the area is
     /// sensible, false otherwise
-    bool OnCPUWrite(VAddr addr, u64 size);
+    bool OnCPUWrite(DAddr addr, u64 size);
 
     /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
-    void FlushAndInvalidateRegion(VAddr addr, u64 size);
+    void FlushAndInvalidateRegion(DAddr addr, u64 size);
 
 private:
     struct Impl;

src/video_core/gpu_thread.cpp

@@ -82,7 +82,7 @@ void ThreadManager::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     PushCommand(SwapBuffersCommand(framebuffer ? std::make_optional(*framebuffer) : std::nullopt));
 }
 
-void ThreadManager::FlushRegion(VAddr addr, u64 size) {
+void ThreadManager::FlushRegion(DAddr addr, u64 size) {
     if (!is_async) {
         // Always flush with synchronous GPU mode
         PushCommand(FlushRegionCommand(addr, size));
@@ -101,11 +101,11 @@ void ThreadManager::TickGPU() {
     PushCommand(GPUTickCommand());
 }
 
-void ThreadManager::InvalidateRegion(VAddr addr, u64 size) {
+void ThreadManager::InvalidateRegion(DAddr addr, u64 size) {
     rasterizer->OnCacheInvalidation(addr, size);
 }
 
-void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
+void ThreadManager::FlushAndInvalidateRegion(DAddr addr, u64 size) {
     // Skip flush on asynch mode, as FlushAndInvalidateRegion is not used for anything too important
     rasterizer->OnCacheInvalidation(addr, size);
 }

src/video_core/gpu_thread.h

@@ -54,26 +54,26 @@ struct SwapBuffersCommand final {
 
 /// Command to signal to the GPU thread to flush a region
 struct FlushRegionCommand final {
-    explicit constexpr FlushRegionCommand(VAddr addr_, u64 size_) : addr{addr_}, size{size_} {}
+    explicit constexpr FlushRegionCommand(DAddr addr_, u64 size_) : addr{addr_}, size{size_} {}
 
-    VAddr addr;
+    DAddr addr;
     u64 size;
 };
 
 /// Command to signal to the GPU thread to invalidate a region
 struct InvalidateRegionCommand final {
-    explicit constexpr InvalidateRegionCommand(VAddr addr_, u64 size_) : addr{addr_}, size{size_} {}
+    explicit constexpr InvalidateRegionCommand(DAddr addr_, u64 size_) : addr{addr_}, size{size_} {}
 
-    VAddr addr;
+    DAddr addr;
     u64 size;
 };
 
 /// Command to signal to the GPU thread to flush and invalidate a region
 struct FlushAndInvalidateRegionCommand final {
-    explicit constexpr FlushAndInvalidateRegionCommand(VAddr addr_, u64 size_)
+    explicit constexpr FlushAndInvalidateRegionCommand(DAddr addr_, u64 size_)
         : addr{addr_}, size{size_} {}
 
-    VAddr addr;
+    DAddr addr;
     u64 size;
 };
 
@@ -122,13 +122,13 @@ public:
     void SwapBuffers(const Tegra::FramebufferConfig* framebuffer);
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    void FlushRegion(VAddr addr, u64 size);
+    void FlushRegion(DAddr addr, u64 size);
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    void InvalidateRegion(VAddr addr, u64 size);
+    void InvalidateRegion(DAddr addr, u64 size);
 
     /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
-    void FlushAndInvalidateRegion(VAddr addr, u64 size);
+    void FlushAndInvalidateRegion(DAddr addr, u64 size);
 
     void TickGPU();
 

src/video_core/guest_memory.h

@@ -0,0 +1,29 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <iterator>
+#include <memory>
+#include <optional>
+#include <span>
+#include <vector>
+
+#include "common/scratch_buffer.h"
+#include "core/guest_memory.h"
+#include "video_core/memory_manager.h"
+
+namespace Tegra::Memory {
+
+using GuestMemoryFlags = Core::Memory::GuestMemoryFlags;
+
+template <typename T, GuestMemoryFlags FLAGS>
+using DeviceGuestMemory = Core::Memory::GuestMemory<Tegra::MaxwellDeviceMemoryManager, T, FLAGS>;
+template <typename T, GuestMemoryFlags FLAGS>
+using DeviceGuestMemoryScoped = Core::Memory::GuestMemoryScoped<Tegra::MaxwellDeviceMemoryManager, T, FLAGS>;
+template <typename T, GuestMemoryFlags FLAGS>
+using GpuGuestMemory = Core::Memory::GuestMemory<Tegra::MemoryManager, T, FLAGS>;
+template <typename T, GuestMemoryFlags FLAGS>
+using GpuGuestMemoryScoped = Core::Memory::GuestMemoryScoped<Tegra::MemoryManager, T, FLAGS>;
+
+} // namespace Tegra::Memory

src/video_core/host1x/gpu_device_memory_manager.h

@@ -1,6 +1,8 @@
 // SPDX-FileCopyrightText: 2023 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
+#pragma once
+
 #include "core/device_memory_manager.h"
 
 namespace VideoCore {
@@ -12,8 +14,8 @@ namespace Tegra {
 struct MaxwellDeviceMethods;
 
 struct MaxwellDeviceTraits {
-    static constexpr bool supports_pinning = true;
+    static constexpr bool supports_pinning = false;
-    static constexpr size_t device_virtual_bits = 34;
+    static constexpr size_t device_virtual_bits = 32;
     using DeviceInterface = typename VideoCore::RasterizerInterface;
     using DeviceMethods = typename MaxwellDeviceMethods;
 };

src/video_core/memory_manager.cpp

@@ -7,22 +7,24 @@
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/device_memory.h"
 #include "core/hle/kernel/k_page_table.h"
 #include "core/hle/kernel/k_process.h"
+#include "video_core/guest_memory.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/invalidation_accumulator.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
 
+
 namespace Tegra {
-using Core::Memory::GuestMemoryFlags;
+using Tegra::Memory::GuestMemoryFlags;
 
 std::atomic<size_t> MemoryManager::unique_identifier_generator{};
 
 MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64 big_page_bits_,
                              u64 page_bits_)
-    : system{system_}, memory{system.ApplicationMemory()}, device_memory{system.DeviceMemory()},
+    : system{system_}, memory{system.Host1x().MemoryManager()},
       address_space_bits{address_space_bits_}, page_bits{page_bits_}, big_page_bits{big_page_bits_},
       entries{}, big_entries{}, page_table{address_space_bits, address_space_bits + page_bits - 38,
                                            page_bits != big_page_bits ? page_bits : 0},
@@ -42,7 +44,7 @@ MemoryManager::MemoryManager(Core::System& system_, u64 address_space_bits_, u64
     big_page_table_mask = big_page_table_size - 1;
 
     big_entries.resize(big_page_table_size / 32, 0);
-    big_page_table_cpu.resize(big_page_table_size);
+    big_page_table_dev.resize(big_page_table_size);
     big_page_continuous.resize(big_page_table_size / continuous_bits, 0);
     entries.resize(page_table_size / 32, 0);
 }
@@ -100,7 +102,7 @@ inline void MemoryManager::SetBigPageContinuous(size_t big_page_index, bool valu
 }
 
 template <MemoryManager::EntryType entry_type>
-GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
+GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size,
                                     PTEKind kind) {
     [[maybe_unused]] u64 remaining_size{size};
     if constexpr (entry_type == EntryType::Mapped) {
@@ -114,9 +116,9 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cp
             rasterizer->ModifyGPUMemory(unique_identifier, current_gpu_addr, page_size);
         }
         if constexpr (entry_type == EntryType::Mapped) {
-            const VAddr current_cpu_addr = cpu_addr + offset;
+            const DAddr current_dev_addr = dev_addr + offset;
             const auto index = PageEntryIndex<false>(current_gpu_addr);
-            const u32 sub_value = static_cast<u32>(current_cpu_addr >> cpu_page_bits);
+            const u32 sub_value = static_cast<u32>(current_dev_addr >> cpu_page_bits);
             page_table[index] = sub_value;
         }
         remaining_size -= page_size;
@@ -126,7 +128,7 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cp
 }
 
 template <MemoryManager::EntryType entry_type>
-GPUVAddr MemoryManager::BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr,
+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) {
@@ -137,20 +139,20 @@ GPUVAddr MemoryManager::BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr
             rasterizer->ModifyGPUMemory(unique_identifier, current_gpu_addr, big_page_size);
         }
         if constexpr (entry_type == EntryType::Mapped) {
-            const VAddr current_cpu_addr = cpu_addr + offset;
+            const DAddr current_dev_addr = dev_addr + offset;
             const auto index = PageEntryIndex<true>(current_gpu_addr);
-            const u32 sub_value = static_cast<u32>(current_cpu_addr >> cpu_page_bits);
+            const u32 sub_value = static_cast<u32>(current_dev_addr >> cpu_page_bits);
-            big_page_table_cpu[index] = sub_value;
+            big_page_table_dev[index] = sub_value;
             const bool is_continuous = ([&] {
                 uintptr_t base_ptr{
-                    reinterpret_cast<uintptr_t>(memory.GetPointerSilent(current_cpu_addr))};
+                    reinterpret_cast<uintptr_t>(memory.GetPointer<u8>(current_dev_addr))};
                 if (base_ptr == 0) {
                     return false;
                 }
-                for (VAddr start_cpu = current_cpu_addr + page_size;
+                for (DAddr start_cpu = current_dev_addr + page_size;
-                     start_cpu < current_cpu_addr + big_page_size; start_cpu += page_size) {
+                     start_cpu < current_dev_addr + big_page_size; start_cpu += page_size) {
                     base_ptr += page_size;
-                    auto next_ptr = reinterpret_cast<uintptr_t>(memory.GetPointerSilent(start_cpu));
+                    auto next_ptr = reinterpret_cast<uintptr_t>(memory.GetPointer<u8>(start_cpu));
                     if (next_ptr == 0 || base_ptr != next_ptr) {
                         return false;
                     }
@@ -172,12 +174,12 @@ void MemoryManager::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_)
     rasterizer = rasterizer_;
 }
 
-GPUVAddr MemoryManager::Map(GPUVAddr gpu_addr, VAddr cpu_addr, std::size_t size, PTEKind kind,
+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, cpu_addr, size, kind);
+        return BigPageTableOp<EntryType::Mapped>(gpu_addr, dev_addr, size, kind);
     }
-    return PageTableOp<EntryType::Mapped>(gpu_addr, cpu_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) {
@@ -202,7 +204,7 @@ void MemoryManager::Unmap(GPUVAddr gpu_addr, std::size_t size) {
     PageTableOp<EntryType::Free>(gpu_addr, 0, size, PTEKind::INVALID);
 }
 
-std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const {
+std::optional<DAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const {
     if (!IsWithinGPUAddressRange(gpu_addr)) [[unlikely]] {
         return std::nullopt;
     }
@@ -211,17 +213,17 @@ std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr gpu_addr) const {
             return std::nullopt;
         }
 
-        const VAddr cpu_addr_base = static_cast<VAddr>(page_table[PageEntryIndex<false>(gpu_addr)])
+        const DAddr dev_addr_base = static_cast<DAddr>(page_table[PageEntryIndex<false>(gpu_addr)])
                                     << cpu_page_bits;
-        return cpu_addr_base + (gpu_addr & page_mask);
+        return dev_addr_base + (gpu_addr & page_mask);
     }
 
-    const VAddr cpu_addr_base =
+    const DAddr dev_addr_base =
-        static_cast<VAddr>(big_page_table_cpu[PageEntryIndex<true>(gpu_addr)]) << cpu_page_bits;
+        static_cast<DAddr>(big_page_table_dev[PageEntryIndex<true>(gpu_addr)]) << cpu_page_bits;
-    return cpu_addr_base + (gpu_addr & big_page_mask);
+    return dev_addr_base + (gpu_addr & big_page_mask);
 }
 
-std::optional<VAddr> MemoryManager::GpuToCpuAddress(GPUVAddr addr, std::size_t size) const {
+std::optional<DAddr> MemoryManager::GpuToCpuAddress(GPUVAddr addr, std::size_t size) const {
     size_t page_index{addr >> page_bits};
     const size_t page_last{(addr + size + page_size - 1) >> page_bits};
     while (page_index < page_last) {
@@ -274,7 +276,7 @@ u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) {
         return {};
     }
 
-    return memory.GetPointer(*address);
+    return memory.GetPointer<u8>(*address);
 }
 
 const u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) const {
@@ -283,7 +285,7 @@ const u8* MemoryManager::GetPointer(GPUVAddr gpu_addr) const {
         return {};
     }
 
-    return memory.GetPointer(*address);
+    return memory.GetPointer<u8>(*address);
 }
 
 #ifdef _MSC_VER // no need for gcc / clang but msvc's compiler is more conservative with inlining.
@@ -367,25 +369,25 @@ void MemoryManager::ReadBlockImpl(GPUVAddr gpu_src_addr, void* dest_buffer, std:
         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 VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
         if constexpr (is_safe) {
-            rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
+            rasterizer->FlushRegion(dev_addr_base, copy_amount, which);
         }
-        u8* physical = memory.GetPointer(cpu_addr_base);
+        u8* physical = memory.GetPointer<u8>(dev_addr_base);
         std::memcpy(dest_buffer, physical, copy_amount);
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
     };
     auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
         if constexpr (is_safe) {
-            rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
+            rasterizer->FlushRegion(dev_addr_base, copy_amount, which);
         }
         if (!IsBigPageContinuous(page_index)) [[unlikely]] {
-            memory.ReadBlockUnsafe(cpu_addr_base, dest_buffer, copy_amount);
+            memory.ReadBlockUnsafe(dev_addr_base, dest_buffer, copy_amount);
         } else {
-            u8* physical = memory.GetPointer(cpu_addr_base);
+            u8* physical = memory.GetPointer<u8>(dev_addr_base);
             std::memcpy(dest_buffer, physical, copy_amount);
         }
         dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
@@ -416,25 +418,25 @@ void MemoryManager::WriteBlockImpl(GPUVAddr gpu_dest_addr, const void* src_buffe
         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 VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
         if constexpr (is_safe) {
-            rasterizer->InvalidateRegion(cpu_addr_base, copy_amount, which);
+            rasterizer->InvalidateRegion(dev_addr_base, copy_amount, which);
         }
-        u8* physical = memory.GetPointer(cpu_addr_base);
+        u8* physical = memory.GetPointer<u8>(dev_addr_base);
         std::memcpy(physical, src_buffer, copy_amount);
         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 VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
         if constexpr (is_safe) {
-            rasterizer->InvalidateRegion(cpu_addr_base, copy_amount, which);
+            rasterizer->InvalidateRegion(dev_addr_base, copy_amount, which);
         }
         if (!IsBigPageContinuous(page_index)) [[unlikely]] {
-            memory.WriteBlockUnsafe(cpu_addr_base, src_buffer, copy_amount);
+            memory.WriteBlockUnsafe(dev_addr_base, src_buffer, copy_amount);
         } else {
-            u8* physical = memory.GetPointer(cpu_addr_base);
+            u8* physical = memory.GetPointer<u8>(dev_addr_base);
             std::memcpy(physical, src_buffer, copy_amount);
         }
         src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
@@ -470,14 +472,14 @@ 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 VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
-        rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
+        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 VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
-        rasterizer->FlushRegion(cpu_addr_base, copy_amount, which);
+        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) {
@@ -495,15 +497,15 @@ bool MemoryManager::IsMemoryDirty(GPUVAddr gpu_addr, size_t size,
                           [[maybe_unused]] std::size_t copy_amount) { return false; };
 
     auto mapped_normal = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
-        result |= rasterizer->MustFlushRegion(cpu_addr_base, copy_amount, which);
+        result |= rasterizer->MustFlushRegion(dev_addr_base, copy_amount, which);
         return result;
     };
     auto mapped_big = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
-        result |= rasterizer->MustFlushRegion(cpu_addr_base, copy_amount, which);
+        result |= rasterizer->MustFlushRegion(dev_addr_base, copy_amount, which);
         return result;
     };
     auto check_short_pages = [&](std::size_t page_index, std::size_t offset,
@@ -517,7 +519,7 @@ bool MemoryManager::IsMemoryDirty(GPUVAddr gpu_addr, size_t size,
 }
 
 size_t MemoryManager::MaxContinuousRange(GPUVAddr gpu_addr, size_t size) const {
-    std::optional<VAddr> old_page_addr{};
+    std::optional<DAddr> old_page_addr{};
     size_t range_so_far = 0;
     bool result{false};
     auto fail = [&]([[maybe_unused]] std::size_t page_index, [[maybe_unused]] std::size_t offset,
@@ -526,24 +528,24 @@ size_t MemoryManager::MaxContinuousRange(GPUVAddr gpu_addr, size_t size) const {
         return true;
     };
     auto short_check = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
-        if (old_page_addr && *old_page_addr != cpu_addr_base) {
+        if (old_page_addr && *old_page_addr != dev_addr_base) {
             result = true;
             return true;
         }
         range_so_far += copy_amount;
-        old_page_addr = {cpu_addr_base + copy_amount};
+        old_page_addr = {dev_addr_base + copy_amount};
         return false;
     };
     auto big_check = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
-        if (old_page_addr && *old_page_addr != cpu_addr_base) {
+        if (old_page_addr && *old_page_addr != dev_addr_base) {
             return true;
         }
         range_so_far += copy_amount;
-        old_page_addr = {cpu_addr_base + copy_amount};
+        old_page_addr = {dev_addr_base + copy_amount};
         return false;
     };
     auto check_short_pages = [&](std::size_t page_index, std::size_t offset,
@@ -568,14 +570,14 @@ 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 VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
-        rasterizer->InvalidateRegion(cpu_addr_base, copy_amount, which);
+        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 VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
-        rasterizer->InvalidateRegion(cpu_addr_base, copy_amount, which);
+        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) {
@@ -587,7 +589,7 @@ void MemoryManager::InvalidateRegion(GPUVAddr gpu_addr, size_t size,
 
 void MemoryManager::CopyBlock(GPUVAddr gpu_dest_addr, GPUVAddr gpu_src_addr, std::size_t size,
                               VideoCommon::CacheType which) {
-    Core::Memory::GpuGuestMemoryScoped<u8, GuestMemoryFlags::SafeReadWrite> data(
+    Tegra::Memory::GpuGuestMemoryScoped<u8, GuestMemoryFlags::SafeReadWrite> data(
         *this, gpu_src_addr, size);
     data.SetAddressAndSize(gpu_dest_addr, size);
     FlushRegion(gpu_dest_addr, size, which);
@@ -611,7 +613,7 @@ bool MemoryManager::IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const {
 }
 
 bool MemoryManager::IsContinuousRange(GPUVAddr gpu_addr, std::size_t size) const {
-    std::optional<VAddr> old_page_addr{};
+    std::optional<DAddr> old_page_addr{};
     bool result{true};
     auto fail = [&]([[maybe_unused]] std::size_t page_index, [[maybe_unused]] std::size_t offset,
                     std::size_t copy_amount) {
@@ -619,23 +621,23 @@ bool MemoryManager::IsContinuousRange(GPUVAddr gpu_addr, std::size_t size) const
         return true;
     };
     auto short_check = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
-        if (old_page_addr && *old_page_addr != cpu_addr_base) {
+        if (old_page_addr && *old_page_addr != dev_addr_base) {
             result = false;
             return true;
         }
-        old_page_addr = {cpu_addr_base + copy_amount};
+        old_page_addr = {dev_addr_base + copy_amount};
         return false;
     };
     auto big_check = [&](std::size_t page_index, std::size_t offset, std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
-        if (old_page_addr && *old_page_addr != cpu_addr_base) {
+        if (old_page_addr && *old_page_addr != dev_addr_base) {
             result = false;
             return true;
         }
-        old_page_addr = {cpu_addr_base + copy_amount};
+        old_page_addr = {dev_addr_base + copy_amount};
         return false;
     };
     auto check_short_pages = [&](std::size_t page_index, std::size_t offset,
@@ -678,11 +680,11 @@ 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, VAddr>, std::size_t>, 32>& result)
+        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, VAddr>, std::size_t>>
+    std::optional<std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>>
         last_segment{};
-    std::optional<VAddr> old_page_addr{};
+    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,
                                                 [[maybe_unused]] std::size_t copy_amount) {
@@ -694,20 +696,20 @@ void MemoryManager::GetSubmappedRangeImpl(
     const auto extend_size_big = [this, &split, &old_page_addr,
                                   &last_segment](std::size_t page_index, std::size_t offset,
                                                  std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(big_page_table_cpu[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(big_page_table_dev[page_index]) << cpu_page_bits) + offset;
         if (old_page_addr) {
-            if (*old_page_addr != cpu_addr_base) {
+            if (*old_page_addr != dev_addr_base) {
                 split(0, 0, 0);
             }
         }
-        old_page_addr = {cpu_addr_base + copy_amount};
+        old_page_addr = {dev_addr_base + copy_amount};
         if (!last_segment) {
             if constexpr (is_gpu_address) {
                 const GPUVAddr new_base_addr = (page_index << big_page_bits) + offset;
                 last_segment = {new_base_addr, copy_amount};
             } else {
-                last_segment = {cpu_addr_base, copy_amount};
+                last_segment = {dev_addr_base, copy_amount};
             }
         } else {
             last_segment->second += copy_amount;
@@ -716,20 +718,20 @@ void MemoryManager::GetSubmappedRangeImpl(
     const auto extend_size_short = [this, &split, &old_page_addr,
                                     &last_segment](std::size_t page_index, std::size_t offset,
                                                    std::size_t copy_amount) {
-        const VAddr cpu_addr_base =
+        const DAddr dev_addr_base =
-            (static_cast<VAddr>(page_table[page_index]) << cpu_page_bits) + offset;
+            (static_cast<DAddr>(page_table[page_index]) << cpu_page_bits) + offset;
         if (old_page_addr) {
-            if (*old_page_addr != cpu_addr_base) {
+            if (*old_page_addr != dev_addr_base) {
                 split(0, 0, 0);
             }
         }
-        old_page_addr = {cpu_addr_base + copy_amount};
+        old_page_addr = {dev_addr_base + copy_amount};
         if (!last_segment) {
             if constexpr (is_gpu_address) {
                 const GPUVAddr new_base_addr = (page_index << page_bits) + offset;
                 last_segment = {new_base_addr, copy_amount};
             } else {
-                last_segment = {cpu_addr_base, copy_amount};
+                last_segment = {dev_addr_base, copy_amount};
             }
         } else {
             last_segment->second += copy_amount;
@@ -756,9 +758,9 @@ void MemoryManager::FlushCaching() {
 }
 
 const u8* MemoryManager::GetSpan(const GPUVAddr src_addr, const std::size_t size) const {
-    auto cpu_addr = GpuToCpuAddress(src_addr);
+    auto dev_addr = GpuToCpuAddress(src_addr);
-    if (cpu_addr) {
+    if (dev_addr) {
-        return memory.GetSpan(*cpu_addr, size);
+        return memory.GetSpan(*dev_addr, size);
     }
     return nullptr;
 }
@@ -767,9 +769,9 @@ u8* MemoryManager::GetSpan(const GPUVAddr src_addr, const std::size_t size) {
     if (!IsContinuousRange(src_addr, size)) {
         return nullptr;
     }
-    auto cpu_addr = GpuToCpuAddress(src_addr);
+    auto dev_addr = GpuToCpuAddress(src_addr);
-    if (cpu_addr) {
+    if (dev_addr) {
-        return memory.GetSpan(*cpu_addr, size);
+        return memory.GetSpan(*dev_addr, size);
     }
     return nullptr;
 }

src/video_core/memory_manager.h

@@ -17,6 +17,7 @@
 #include "common/virtual_buffer.h"
 #include "core/memory.h"
 #include "video_core/cache_types.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/pte_kind.h"
 
 namespace VideoCore {
@@ -28,10 +29,6 @@ class InvalidationAccumulator;
 }
 
 namespace Core {
-class DeviceMemory;
-namespace Memory {
-class Memory;
-} // namespace Memory
 class System;
 } // namespace Core
 
@@ -50,9 +47,9 @@ public:
     /// Binds a renderer to the memory manager.
     void BindRasterizer(VideoCore::RasterizerInterface* rasterizer);
 
-    [[nodiscard]] std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr) const;
+    [[nodiscard]] std::optional<DAddr> GpuToCpuAddress(GPUVAddr addr) const;
 
-    [[nodiscard]] std::optional<VAddr> GpuToCpuAddress(GPUVAddr addr, std::size_t size) const;
+    [[nodiscard]] std::optional<DAddr> GpuToCpuAddress(GPUVAddr addr, std::size_t size) const;
 
     template <typename T>
     [[nodiscard]] T Read(GPUVAddr addr) const;
@@ -110,7 +107,7 @@ public:
     [[nodiscard]] bool IsGranularRange(GPUVAddr gpu_addr, std::size_t size) const;
 
     /**
-     * Checks if a gpu region is mapped by a single range of cpu addresses.
+     * Checks if a gpu region is mapped by a single range of device addresses.
      */
     [[nodiscard]] bool IsContinuousRange(GPUVAddr gpu_addr, std::size_t size) const;
 
@@ -120,14 +117,14 @@ public:
     [[nodiscard]] bool IsFullyMappedRange(GPUVAddr gpu_addr, std::size_t size) const;
 
     /**
-     * Returns a vector with all the subranges of cpu addresses mapped beneath.
+     * Returns a vector with all the subranges of device addresses mapped beneath.
      * if the region is continuous, a single pair will be returned. If it's unmapped, an empty
      * vector will be returned;
      */
     boost::container::small_vector<std::pair<GPUVAddr, std::size_t>, 32> GetSubmappedRange(
         GPUVAddr gpu_addr, std::size_t size) const;
 
-    GPUVAddr Map(GPUVAddr gpu_addr, VAddr cpu_addr, std::size_t size,
+    GPUVAddr Map(GPUVAddr gpu_addr, DAddr dev_addr, std::size_t size,
                  PTEKind kind = PTEKind::INVALID, bool is_big_pages = true);
     GPUVAddr MapSparse(GPUVAddr gpu_addr, std::size_t size, bool is_big_pages = true);
     void Unmap(GPUVAddr gpu_addr, std::size_t size);
@@ -186,12 +183,11 @@ private:
     void GetSubmappedRangeImpl(
         GPUVAddr gpu_addr, std::size_t size,
         boost::container::small_vector<
-            std::pair<std::conditional_t<is_gpu_address, GPUVAddr, VAddr>, std::size_t>, 32>&
+            std::pair<std::conditional_t<is_gpu_address, GPUVAddr, DAddr>, std::size_t>, 32>&
             result) const;
 
     Core::System& system;
-    Core::Memory::Memory& memory;
+    MaxwellDeviceMemoryManager& memory;
-    Core::DeviceMemory& device_memory;
 
     const u64 address_space_bits;
     const u64 page_bits;
@@ -218,11 +214,11 @@ private:
     std::vector<u64> big_entries;
 
     template <EntryType entry_type>
-    GPUVAddr PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
+    GPUVAddr PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size,
                          PTEKind kind);
 
     template <EntryType entry_type>
-    GPUVAddr BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
+    GPUVAddr BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] DAddr dev_addr, size_t size,
                             PTEKind kind);
 
     template <bool is_big_page>
@@ -233,11 +229,11 @@ private:
 
     Common::MultiLevelPageTable<u32> page_table;
     Common::RangeMap<GPUVAddr, PTEKind> kind_map;
-    Common::VirtualBuffer<u32> big_page_table_cpu;
+    Common::VirtualBuffer<u32> big_page_table_dev;
 
     std::vector<u64> big_page_continuous;
-    boost::container::small_vector<std::pair<VAddr, std::size_t>, 32> page_stash{};
+    boost::container::small_vector<std::pair<DAddr, std::size_t>, 32> page_stash{};
-    boost::container::small_vector<std::pair<VAddr, std::size_t>, 32> page_stash2{};
+    boost::container::small_vector<std::pair<DAddr, std::size_t>, 32> page_stash2{};
 
     mutable std::mutex guard;
 

src/video_core/query_cache.h

@@ -21,6 +21,7 @@
 #include "core/memory.h"
 #include "video_core/control/channel_state_cache.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/texture_cache/slot_vector.h"
@@ -102,18 +103,19 @@ template <class QueryCache, class CachedQuery, class CounterStream, class HostCo
 class QueryCacheLegacy : public VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo> {
 public:
     explicit QueryCacheLegacy(VideoCore::RasterizerInterface& rasterizer_,
-                              Core::Memory::Memory& cpu_memory_)
+                              Tegra::MaxwellDeviceMemoryManager& device_memory_)
         : rasterizer{rasterizer_},
           // Use reinterpret_cast instead of static_cast as workaround for
           // UBSan bug (https://github.com/llvm/llvm-project/issues/59060)
-          cpu_memory{cpu_memory_}, streams{{
+          device_memory{device_memory_},
-                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+          streams{{
-                                                      VideoCore::QueryType::SamplesPassed}},
+              {CounterStream{reinterpret_cast<QueryCache&>(*this),
-                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                             VideoCore::QueryType::SamplesPassed}},
-                                                      VideoCore::QueryType::PrimitivesGenerated}},
+              {CounterStream{reinterpret_cast<QueryCache&>(*this),
-                                       {CounterStream{reinterpret_cast<QueryCache&>(*this),
+                             VideoCore::QueryType::PrimitivesGenerated}},
-                                                      VideoCore::QueryType::TfbPrimitivesWritten}},
+              {CounterStream{reinterpret_cast<QueryCache&>(*this),
-                                   }} {
+                             VideoCore::QueryType::TfbPrimitivesWritten}},
+          }} {
         (void)slot_async_jobs.insert(); // Null value
     }
 
@@ -322,13 +324,14 @@ private:
             local_lock.unlock();
             if (timestamp) {
                 u64 timestamp_value = *timestamp;
-                cpu_memory.WriteBlockUnsafe(address + sizeof(u64), &timestamp_value, sizeof(u64));
+                device_memory.WriteBlockUnsafe(address + sizeof(u64), &timestamp_value,
-                cpu_memory.WriteBlockUnsafe(address, &value, sizeof(u64));
+                                               sizeof(u64));
+                device_memory.WriteBlockUnsafe(address, &value, sizeof(u64));
                 rasterizer.InvalidateRegion(address, sizeof(u64) * 2,
                                             VideoCommon::CacheType::NoQueryCache);
             } else {
                 u32 small_value = static_cast<u32>(value);
-                cpu_memory.WriteBlockUnsafe(address, &small_value, sizeof(u32));
+                device_memory.WriteBlockUnsafe(address, &small_value, sizeof(u32));
                 rasterizer.InvalidateRegion(address, sizeof(u32),
                                             VideoCommon::CacheType::NoQueryCache);
             }
@@ -342,7 +345,7 @@ private:
     SlotVector<AsyncJob> slot_async_jobs;
 
     VideoCore::RasterizerInterface& rasterizer;
-    Core::Memory::Memory& cpu_memory;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
 
     mutable std::recursive_mutex mutex;
 

src/video_core/query_cache/query_base.h

@@ -23,7 +23,7 @@ DECLARE_ENUM_FLAG_OPERATORS(QueryFlagBits)
 
 class QueryBase {
 public:
-    VAddr guest_address{};
+    DAddr guest_address{};
     QueryFlagBits flags{};
     u64 value{};
 
@@ -32,7 +32,7 @@ protected:
     QueryBase() = default;
 
     // Parameterized constructor
-    QueryBase(VAddr address, QueryFlagBits flags_, u64 value_)
+    QueryBase(DAddr address, QueryFlagBits flags_, u64 value_)
         : guest_address(address), flags(flags_), value{value_} {}
 };
 

src/video_core/query_cache/query_cache.h

@@ -18,6 +18,7 @@
 #include "core/memory.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/gpu.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/memory_manager.h"
 #include "video_core/query_cache/bank_base.h"
 #include "video_core/query_cache/query_base.h"
@@ -113,9 +114,10 @@ struct QueryCacheBase<Traits>::QueryCacheBaseImpl {
     using RuntimeType = typename Traits::RuntimeType;
 
     QueryCacheBaseImpl(QueryCacheBase<Traits>* owner_, VideoCore::RasterizerInterface& rasterizer_,
-                       Core::Memory::Memory& cpu_memory_, RuntimeType& runtime_, Tegra::GPU& gpu_)
+                       Tegra::MaxwellDeviceMemoryManager& device_memory_, RuntimeType& runtime_,
+                       Tegra::GPU& gpu_)
         : owner{owner_}, rasterizer{rasterizer_},
-          cpu_memory{cpu_memory_}, runtime{runtime_}, gpu{gpu_} {
+          device_memory{device_memory_}, runtime{runtime_}, gpu{gpu_} {
         streamer_mask = 0;
         for (size_t i = 0; i < static_cast<size_t>(QueryType::MaxQueryTypes); i++) {
             streamers[i] = runtime.GetStreamerInterface(static_cast<QueryType>(i));
@@ -158,7 +160,7 @@ struct QueryCacheBase<Traits>::QueryCacheBaseImpl {
 
     QueryCacheBase<Traits>* owner;
     VideoCore::RasterizerInterface& rasterizer;
-    Core::Memory::Memory& cpu_memory;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
     RuntimeType& runtime;
     Tegra::GPU& gpu;
     std::array<StreamerInterface*, static_cast<size_t>(QueryType::MaxQueryTypes)> streamers;
@@ -171,10 +173,11 @@ struct QueryCacheBase<Traits>::QueryCacheBaseImpl {
 template <typename Traits>
 QueryCacheBase<Traits>::QueryCacheBase(Tegra::GPU& gpu_,
                                        VideoCore::RasterizerInterface& rasterizer_,
-                                       Core::Memory::Memory& cpu_memory_, RuntimeType& runtime_)
+                                       Tegra::MaxwellDeviceMemoryManager& device_memory_,
+                                       RuntimeType& runtime_)
     : cached_queries{} {
     impl = std::make_unique<QueryCacheBase<Traits>::QueryCacheBaseImpl>(
-        this, rasterizer_, cpu_memory_, runtime_, gpu_);
+        this, rasterizer_, device_memory_, runtime_, gpu_);
 }
 
 template <typename Traits>
@@ -240,7 +243,7 @@ void QueryCacheBase<Traits>::CounterReport(GPUVAddr addr, QueryType counter_type
     if (!cpu_addr_opt) [[unlikely]] {
         return;
     }
-    VAddr cpu_addr = *cpu_addr_opt;
+    DAddr cpu_addr = *cpu_addr_opt;
     const size_t new_query_id = streamer->WriteCounter(cpu_addr, has_timestamp, payload, subreport);
     auto* query = streamer->GetQuery(new_query_id);
     if (is_fence) {
@@ -253,10 +256,9 @@ void QueryCacheBase<Traits>::CounterReport(GPUVAddr addr, QueryType counter_type
         return std::make_pair<u64, u32>(cur_addr >> Core::Memory::YUZU_PAGEBITS,
                                         static_cast<u32>(cur_addr & Core::Memory::YUZU_PAGEMASK));
     };
-    u8* pointer = impl->cpu_memory.GetPointer(cpu_addr);
+    u8* pointer = impl->device_memory.GetPointer<u8>(cpu_addr);
-    u8* pointer_timestamp = impl->cpu_memory.GetPointer(cpu_addr + 8);
+    u8* pointer_timestamp = impl->device_memory.GetPointer<u8>(cpu_addr + 8);
     bool is_synced = !Settings::IsGPULevelHigh() && is_fence;
-
     std::function<void()> operation([this, is_synced, streamer, query_base = query, query_location,
                                      pointer, pointer_timestamp] {
         if (True(query_base->flags & QueryFlagBits::IsInvalidated)) {
@@ -559,7 +561,7 @@ bool QueryCacheBase<Traits>::SemiFlushQueryDirty(QueryCacheBase<Traits>::QueryLo
     }
     if (True(query_base->flags & QueryFlagBits::IsFinalValueSynced) &&
         False(query_base->flags & QueryFlagBits::IsGuestSynced)) {
-        auto* ptr = impl->cpu_memory.GetPointer(query_base->guest_address);
+        auto* ptr = impl->device_memory.GetPointer<u8>(query_base->guest_address);
         if (True(query_base->flags & QueryFlagBits::HasTimestamp)) {
             std::memcpy(ptr, &query_base->value, sizeof(query_base->value));
             return false;

src/video_core/query_cache/query_cache_base.h

@@ -17,10 +17,7 @@
 #include "video_core/control/channel_state_cache.h"
 #include "video_core/query_cache/query_base.h"
 #include "video_core/query_cache/types.h"
-
+#include "video_core/host1x/gpu_device_memory_manager.h"
-namespace Core::Memory {
-class Memory;
-}
 
 namespace VideoCore {
 class RasterizerInterface;
@@ -53,7 +50,7 @@ public:
     };
 
     explicit QueryCacheBase(Tegra::GPU& gpu, VideoCore::RasterizerInterface& rasterizer_,
-                            Core::Memory::Memory& cpu_memory_, RuntimeType& runtime_);
+                            Tegra::MaxwellDeviceMemoryManager& device_memory_, RuntimeType& runtime_);
 
     ~QueryCacheBase();
 

src/video_core/rasterizer_accelerated.cpp

@@ -1,72 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#include <atomic>
-
-#include "common/assert.h"
-#include "common/common_types.h"
-#include "common/div_ceil.h"
-#include "core/memory.h"
-#include "video_core/rasterizer_accelerated.h"
-
-namespace VideoCore {
-
-using namespace Core::Memory;
-
-RasterizerAccelerated::RasterizerAccelerated(Memory& cpu_memory_)
-    : cached_pages(std::make_unique<CachedPages>()), cpu_memory{cpu_memory_} {}
-
-RasterizerAccelerated::~RasterizerAccelerated() = default;
-
-void RasterizerAccelerated::UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {
-    u64 uncache_begin = 0;
-    u64 cache_begin = 0;
-    u64 uncache_bytes = 0;
-    u64 cache_bytes = 0;
-
-    std::atomic_thread_fence(std::memory_order_acquire);
-    const u64 page_end = Common::DivCeil(addr + size, YUZU_PAGESIZE);
-    for (u64 page = addr >> YUZU_PAGEBITS; page != page_end; ++page) {
-        std::atomic_uint16_t& count = cached_pages->at(page >> 2).Count(page);
-
-        if (delta > 0) {
-            ASSERT_MSG(count.load(std::memory_order::relaxed) < UINT16_MAX, "Count may overflow!");
-        } else if (delta < 0) {
-            ASSERT_MSG(count.load(std::memory_order::relaxed) > 0, "Count may underflow!");
-        } else {
-            ASSERT_MSG(false, "Delta must be non-zero!");
-        }
-
-        // Adds or subtracts 1, as count is a unsigned 8-bit value
-        count.fetch_add(static_cast<u16>(delta), std::memory_order_release);
-
-        // Assume delta is either -1 or 1
-        if (count.load(std::memory_order::relaxed) == 0) {
-            if (uncache_bytes == 0) {
-                uncache_begin = page;
-            }
-            uncache_bytes += YUZU_PAGESIZE;
-        } else if (uncache_bytes > 0) {
-            cpu_memory.RasterizerMarkRegionCached(uncache_begin << YUZU_PAGEBITS, uncache_bytes,
-                                                  false);
-            uncache_bytes = 0;
-        }
-        if (count.load(std::memory_order::relaxed) == 1 && delta > 0) {
-            if (cache_bytes == 0) {
-                cache_begin = page;
-            }
-            cache_bytes += YUZU_PAGESIZE;
-        } else if (cache_bytes > 0) {
-            cpu_memory.RasterizerMarkRegionCached(cache_begin << YUZU_PAGEBITS, cache_bytes, true);
-            cache_bytes = 0;
-        }
-    }
-    if (uncache_bytes > 0) {
-        cpu_memory.RasterizerMarkRegionCached(uncache_begin << YUZU_PAGEBITS, uncache_bytes, false);
-    }
-    if (cache_bytes > 0) {
-        cpu_memory.RasterizerMarkRegionCached(cache_begin << YUZU_PAGEBITS, cache_bytes, true);
-    }
-}
-
-} // namespace VideoCore

src/video_core/rasterizer_accelerated.h

@@ -1,49 +0,0 @@
-// SPDX-FileCopyrightText: Copyright 2019 yuzu Emulator Project
-// SPDX-License-Identifier: GPL-2.0-or-later
-
-#pragma once
-
-#include <array>
-#include <atomic>
-
-#include "common/common_types.h"
-#include "video_core/rasterizer_interface.h"
-
-namespace Core::Memory {
-class Memory;
-}
-
-namespace VideoCore {
-
-/// Implements the shared part in GPU accelerated rasterizers in RasterizerInterface.
-class RasterizerAccelerated : public RasterizerInterface {
-public:
-    explicit RasterizerAccelerated(Core::Memory::Memory& cpu_memory_);
-    ~RasterizerAccelerated() override;
-
-    void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) override;
-
-private:
-    class CacheEntry final {
-    public:
-        CacheEntry() = default;
-
-        std::atomic_uint16_t& Count(std::size_t page) {
-            return values[page & 3];
-        }
-
-        const std::atomic_uint16_t& Count(std::size_t page) const {
-            return values[page & 3];
-        }
-
-    private:
-        std::array<std::atomic_uint16_t, 4> values{};
-    };
-    static_assert(sizeof(CacheEntry) == 8, "CacheEntry should be 8 bytes!");
-
-    using CachedPages = std::array<CacheEntry, 0x2000000>;
-    std::unique_ptr<CachedPages> cached_pages;
-    Core::Memory::Memory& cpu_memory;
-};
-
-} // namespace VideoCore

src/video_core/rasterizer_interface.h

@@ -86,35 +86,35 @@ public:
     virtual void FlushAll() = 0;
 
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
-    virtual void FlushRegion(VAddr addr, u64 size,
+    virtual void FlushRegion(DAddr addr, u64 size,
                              VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
     /// Check if the the specified memory area requires flushing to CPU Memory.
-    virtual bool MustFlushRegion(VAddr addr, u64 size,
+    virtual bool MustFlushRegion(DAddr addr, u64 size,
                                  VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
-    virtual RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) = 0;
+    virtual RasterizerDownloadArea GetFlushArea(DAddr addr, u64 size) = 0;
 
     /// Notify rasterizer that any caches of the specified region should be invalidated
-    virtual void InvalidateRegion(VAddr addr, u64 size,
+    virtual void InvalidateRegion(DAddr addr, u64 size,
                                   VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
-    virtual void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) {
+    virtual void InnerInvalidation(std::span<const std::pair<DAddr, std::size_t>> sequences) {
         for (const auto& [cpu_addr, size] : sequences) {
             InvalidateRegion(cpu_addr, size);
         }
     }
 
     /// Notify rasterizer that any caches of the specified region are desync with guest
-    virtual void OnCacheInvalidation(VAddr addr, u64 size) = 0;
+    virtual void OnCacheInvalidation(PAddr addr, u64 size) = 0;
 
-    virtual bool OnCPUWrite(VAddr addr, u64 size) = 0;
+    virtual bool OnCPUWrite(PAddr addr, u64 size) = 0;
 
     /// Sync memory between guest and host.
     virtual void InvalidateGPUCache() = 0;
 
     /// Unmap memory range
-    virtual void UnmapMemory(VAddr addr, u64 size) = 0;
+    virtual void UnmapMemory(DAddr addr, u64 size) = 0;
 
     /// Remap GPU memory range. This means underneath backing memory changed
     virtual void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) = 0;
@@ -122,7 +122,7 @@ public:
     /// Notify rasterizer that any caches of the specified region should be flushed to Switch memory
     /// and invalidated
     virtual void FlushAndInvalidateRegion(
-        VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
+        DAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) = 0;
 
     /// Notify the host renderer to wait for previous primitive and compute operations.
     virtual void WaitForIdle() = 0;
@@ -157,13 +157,10 @@ public:
 
     /// Attempt to use a faster method to display the framebuffer to screen
     [[nodiscard]] virtual bool AccelerateDisplay(const Tegra::FramebufferConfig& config,
-                                                 VAddr framebuffer_addr, u32 pixel_stride) {
+                                                 DAddr framebuffer_addr, u32 pixel_stride) {
         return false;
     }
 
-    /// Increase/decrease the number of object in pages touching the specified region
-    virtual void UpdatePagesCachedCount(VAddr addr, u64 size, int delta) {}
-
     /// Initialize disk cached resources for the game being emulated
     virtual void LoadDiskResources(u64 title_id, std::stop_token stop_loading,
                                    const DiskResourceLoadCallback& callback) {}

src/video_core/renderer_null/null_rasterizer.cpp

@@ -19,8 +19,7 @@ bool AccelerateDMA::BufferClear(GPUVAddr src_address, u64 amount, u32 value) {
     return true;
 }
 
-RasterizerNull::RasterizerNull(Core::Memory::Memory& cpu_memory_, Tegra::GPU& gpu)
+RasterizerNull::RasterizerNull(Tegra::GPU& gpu) : m_gpu{gpu} {}
-    : RasterizerAccelerated(cpu_memory_), m_gpu{gpu} {}
 RasterizerNull::~RasterizerNull() = default;
 
 void RasterizerNull::Draw(bool is_indexed, u32 instance_count) {}
@@ -45,16 +44,16 @@ void RasterizerNull::BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr
                                                u32 size) {}
 void RasterizerNull::DisableGraphicsUniformBuffer(size_t stage, u32 index) {}
 void RasterizerNull::FlushAll() {}
-void RasterizerNull::FlushRegion(VAddr addr, u64 size, VideoCommon::CacheType) {}
+void RasterizerNull::FlushRegion(DAddr addr, u64 size, VideoCommon::CacheType) {}
-bool RasterizerNull::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheType) {
+bool RasterizerNull::MustFlushRegion(DAddr addr, u64 size, VideoCommon::CacheType) {
     return false;
 }
-void RasterizerNull::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType) {}
+void RasterizerNull::InvalidateRegion(DAddr addr, u64 size, VideoCommon::CacheType) {}
-bool RasterizerNull::OnCPUWrite(VAddr addr, u64 size) {
+bool RasterizerNull::OnCPUWrite(PAddr addr, u64 size) {
     return false;
 }
-void RasterizerNull::OnCacheInvalidation(VAddr addr, u64 size) {}
+void RasterizerNull::OnCacheInvalidation(PAddr addr, u64 size) {}
-VideoCore::RasterizerDownloadArea RasterizerNull::GetFlushArea(VAddr addr, u64 size) {
+VideoCore::RasterizerDownloadArea RasterizerNull::GetFlushArea(PAddr addr, u64 size) {
     VideoCore::RasterizerDownloadArea new_area{
         .start_address = Common::AlignDown(addr, Core::Memory::YUZU_PAGESIZE),
         .end_address = Common::AlignUp(addr + size, Core::Memory::YUZU_PAGESIZE),
@@ -63,7 +62,7 @@ VideoCore::RasterizerDownloadArea RasterizerNull::GetFlushArea(VAddr addr, u64 s
     return new_area;
 }
 void RasterizerNull::InvalidateGPUCache() {}
-void RasterizerNull::UnmapMemory(VAddr addr, u64 size) {}
+void RasterizerNull::UnmapMemory(DAddr addr, u64 size) {}
 void RasterizerNull::ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) {}
 void RasterizerNull::SignalFence(std::function<void()>&& func) {
     func();
@@ -78,7 +77,7 @@ void RasterizerNull::SignalSyncPoint(u32 value) {
 }
 void RasterizerNull::SignalReference() {}
 void RasterizerNull::ReleaseFences(bool) {}
-void RasterizerNull::FlushAndInvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType) {}
+void RasterizerNull::FlushAndInvalidateRegion(DAddr addr, u64 size, VideoCommon::CacheType) {}
 void RasterizerNull::WaitForIdle() {}
 void RasterizerNull::FragmentBarrier() {}
 void RasterizerNull::TiledCacheBarrier() {}
@@ -95,7 +94,7 @@ bool RasterizerNull::AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Surfac
 void RasterizerNull::AccelerateInlineToMemory(GPUVAddr address, size_t copy_size,
                                               std::span<const u8> memory) {}
 bool RasterizerNull::AccelerateDisplay(const Tegra::FramebufferConfig& config,
-                                       VAddr framebuffer_addr, u32 pixel_stride) {
+                                       DAddr framebuffer_addr, u32 pixel_stride) {
     return true;
 }
 void RasterizerNull::LoadDiskResources(u64 title_id, std::stop_token stop_loading,

src/video_core/renderer_null/null_rasterizer.h

@@ -6,7 +6,6 @@
 #include "common/common_types.h"
 #include "video_core/control/channel_state_cache.h"
 #include "video_core/engines/maxwell_dma.h"
-#include "video_core/rasterizer_accelerated.h"
 #include "video_core/rasterizer_interface.h"
 
 namespace Core {
@@ -32,10 +31,10 @@ public:
     }
 };
 
-class RasterizerNull final : public VideoCore::RasterizerAccelerated,
+class RasterizerNull final : public VideoCore::RasterizerInterface,
                              protected VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo> {
 public:
-    explicit RasterizerNull(Core::Memory::Memory& cpu_memory, Tegra::GPU& gpu);
+    explicit RasterizerNull(Tegra::GPU& gpu);
     ~RasterizerNull() override;
 
     void Draw(bool is_indexed, u32 instance_count) override;
@@ -48,17 +47,17 @@ public:
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
     void DisableGraphicsUniformBuffer(size_t stage, u32 index) override;
     void FlushAll() override;
-    void FlushRegion(VAddr addr, u64 size,
+    void FlushRegion(DAddr addr, u64 size,
                      VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    bool MustFlushRegion(VAddr addr, u64 size,
+    bool MustFlushRegion(DAddr addr, u64 size,
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    void InvalidateRegion(VAddr addr, u64 size,
+    void InvalidateRegion(DAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    void OnCacheInvalidation(VAddr addr, u64 size) override;
+    void OnCacheInvalidation(DAddr addr, u64 size) override;
-    bool OnCPUWrite(VAddr addr, u64 size) override;
+    bool OnCPUWrite(DAddr addr, u64 size) override;
-    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(DAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
-    void UnmapMemory(VAddr addr, u64 size) override;
+    void UnmapMemory(DAddr addr, u64 size) override;
     void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) override;
     void SignalFence(std::function<void()>&& func) override;
     void SyncOperation(std::function<void()>&& func) override;
@@ -66,7 +65,7 @@ public:
     void SignalReference() override;
     void ReleaseFences(bool force) override;
     void FlushAndInvalidateRegion(
-        VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+        DAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void WaitForIdle() override;
     void FragmentBarrier() override;
     void TiledCacheBarrier() override;
@@ -78,7 +77,7 @@ public:
     Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() override;
     void AccelerateInlineToMemory(GPUVAddr address, size_t copy_size,
                                   std::span<const u8> memory) override;
-    bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
+    bool AccelerateDisplay(const Tegra::FramebufferConfig& config, DAddr framebuffer_addr,
                            u32 pixel_stride) override;
     void LoadDiskResources(u64 title_id, std::stop_token stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;

src/video_core/renderer_null/renderer_null.cpp

@@ -7,10 +7,9 @@
 
 namespace Null {
 
-RendererNull::RendererNull(Core::Frontend::EmuWindow& emu_window, Core::Memory::Memory& cpu_memory,
+RendererNull::RendererNull(Core::Frontend::EmuWindow& emu_window, Tegra::GPU& gpu,
-                           Tegra::GPU& gpu,
                            std::unique_ptr<Core::Frontend::GraphicsContext> context_)
-    : RendererBase(emu_window, std::move(context_)), m_gpu(gpu), m_rasterizer(cpu_memory, gpu) {}
+    : RendererBase(emu_window, std::move(context_)), m_gpu(gpu), m_rasterizer(gpu) {}
 
 RendererNull::~RendererNull() = default;
 

src/video_core/renderer_null/renderer_null.h

@@ -13,8 +13,7 @@ namespace Null {
 
 class RendererNull final : public VideoCore::RendererBase {
 public:
-    explicit RendererNull(Core::Frontend::EmuWindow& emu_window, Core::Memory::Memory& cpu_memory,
+    explicit RendererNull(Core::Frontend::EmuWindow& emu_window, Tegra::GPU& gpu,
-                          Tegra::GPU& gpu,
                           std::unique_ptr<Core::Frontend::GraphicsContext> context);
     ~RendererNull() override;
 

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -47,11 +47,10 @@ constexpr std::array PROGRAM_LUT{
 } // Anonymous namespace
 
 Buffer::Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params)
-    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params) {}
+    : VideoCommon::BufferBase(null_params) {}
 
-Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
+Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_)
-               VAddr cpu_addr_, u64 size_bytes_)
+    : VideoCommon::BufferBase(cpu_addr_, size_bytes_) {
-    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(rasterizer_, cpu_addr_, size_bytes_) {
     buffer.Create();
     if (runtime.device.HasDebuggingToolAttached()) {
         const std::string name = fmt::format("Buffer 0x{:x}", CpuAddr());

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -10,7 +10,6 @@
 #include "common/common_types.h"
 #include "video_core/buffer_cache/buffer_cache_base.h"
 #include "video_core/buffer_cache/memory_tracker_base.h"
-#include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_staging_buffer_pool.h"
@@ -19,9 +18,9 @@ namespace OpenGL {
 
 class BufferCacheRuntime;
 
-class Buffer : public VideoCommon::BufferBase<VideoCore::RasterizerInterface> {
+class Buffer : public VideoCommon::BufferBase {
 public:
-    explicit Buffer(BufferCacheRuntime&, VideoCore::RasterizerInterface& rasterizer, VAddr cpu_addr,
+    explicit Buffer(BufferCacheRuntime&, DAddr cpu_addr,
                     u64 size_bytes);
     explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams);
 
@@ -244,7 +243,7 @@ struct BufferCacheParams {
     using Runtime = OpenGL::BufferCacheRuntime;
     using Buffer = OpenGL::Buffer;
     using Async_Buffer = OpenGL::StagingBufferMap;
-    using MemoryTracker = VideoCommon::MemoryTrackerBase<VideoCore::RasterizerInterface>;
+    using MemoryTracker = VideoCommon::MemoryTrackerBase<Tegra::MaxwellDeviceMemoryManager>;
 
     static constexpr bool IS_OPENGL = true;
     static constexpr bool HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS = true;

src/video_core/renderer_opengl/gl_query_cache.cpp

@@ -35,8 +35,8 @@ constexpr GLenum GetTarget(VideoCore::QueryType type) {
 
 } // Anonymous namespace
 
-QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_)
+QueryCache::QueryCache(RasterizerOpenGL& rasterizer_, Tegra::MaxwellDeviceMemoryManager& device_memory_)
-    : QueryCacheLegacy(rasterizer_, cpu_memory_), gl_rasterizer{rasterizer_} {
+    : QueryCacheLegacy(rasterizer_, device_memory_), gl_rasterizer{rasterizer_} {
     EnableCounters();
 }
 

src/video_core/renderer_opengl/gl_query_cache.h

@@ -11,6 +11,7 @@
 #include "video_core/query_cache.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 
 namespace Core {
 class System;
@@ -28,7 +29,7 @@ using CounterStream = VideoCommon::CounterStreamBase<QueryCache, HostCounter>;
 class QueryCache final
     : public VideoCommon::QueryCacheLegacy<QueryCache, CachedQuery, CounterStream, HostCounter> {
 public:
-    explicit QueryCache(RasterizerOpenGL& rasterizer_, Core::Memory::Memory& cpu_memory_);
+    explicit QueryCache(RasterizerOpenGL& rasterizer_, Tegra::MaxwellDeviceMemoryManager& device_memory_);
     ~QueryCache();
 
     OGLQuery AllocateQuery(VideoCore::QueryType type);

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -70,18 +70,18 @@ std::optional<VideoCore::QueryType> MaxwellToVideoCoreQuery(VideoCommon::QueryTy
 } // Anonymous namespace
 
 RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
-                                   Core::Memory::Memory& cpu_memory_, const Device& device_,
+                                   Tegra::MaxwellDeviceMemoryManager& device_memory_,
-                                   ScreenInfo& screen_info_, ProgramManager& program_manager_,
+                                   const Device& device_, ScreenInfo& screen_info_,
-                                   StateTracker& state_tracker_)
+                                   ProgramManager& program_manager_, StateTracker& state_tracker_)
-    : RasterizerAccelerated(cpu_memory_), gpu(gpu_), device(device_), screen_info(screen_info_),
+    : gpu(gpu_), device_memory(device_memory_), device(device_), screen_info(screen_info_),
       program_manager(program_manager_), state_tracker(state_tracker_),
       texture_cache_runtime(device, program_manager, state_tracker, staging_buffer_pool),
-      texture_cache(texture_cache_runtime, *this),
+      texture_cache(texture_cache_runtime, device_memory_),
       buffer_cache_runtime(device, staging_buffer_pool),
-      buffer_cache(*this, cpu_memory_, buffer_cache_runtime),
+      buffer_cache(device_memory_, buffer_cache_runtime),
-      shader_cache(*this, emu_window_, device, texture_cache, buffer_cache, program_manager,
+      shader_cache(device_memory_, emu_window_, device, texture_cache, buffer_cache,
-                   state_tracker, gpu.ShaderNotify()),
+                   program_manager, state_tracker, gpu.ShaderNotify()),
-      query_cache(*this, cpu_memory_), accelerate_dma(buffer_cache, texture_cache),
+      query_cache(*this, device_memory_), accelerate_dma(buffer_cache, texture_cache),
       fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache),
       blit_image(program_manager_) {}
 
@@ -475,7 +475,7 @@ void RasterizerOpenGL::DisableGraphicsUniformBuffer(size_t stage, u32 index) {
 
 void RasterizerOpenGL::FlushAll() {}
 
-void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
+void RasterizerOpenGL::FlushRegion(DAddr addr, u64 size, VideoCommon::CacheType which) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     if (addr == 0 || size == 0) {
         return;
@@ -493,7 +493,7 @@ void RasterizerOpenGL::FlushRegion(VAddr addr, u64 size, VideoCommon::CacheType
     }
 }
 
-bool RasterizerOpenGL::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
+bool RasterizerOpenGL::MustFlushRegion(DAddr addr, u64 size, VideoCommon::CacheType which) {
     if ((True(which & VideoCommon::CacheType::BufferCache))) {
         std::scoped_lock lock{buffer_cache.mutex};
         if (buffer_cache.IsRegionGpuModified(addr, size)) {
@@ -510,7 +510,7 @@ bool RasterizerOpenGL::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheT
     return false;
 }
 
-VideoCore::RasterizerDownloadArea RasterizerOpenGL::GetFlushArea(VAddr addr, u64 size) {
+VideoCore::RasterizerDownloadArea RasterizerOpenGL::GetFlushArea(DAddr addr, u64 size) {
     {
         std::scoped_lock lock{texture_cache.mutex};
         auto area = texture_cache.GetFlushArea(addr, size);
@@ -533,7 +533,7 @@ VideoCore::RasterizerDownloadArea RasterizerOpenGL::GetFlushArea(VAddr addr, u64
     return new_area;
 }
 
-void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
+void RasterizerOpenGL::InvalidateRegion(DAddr addr, u64 size, VideoCommon::CacheType which) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
     if (addr == 0 || size == 0) {
         return;
@@ -554,8 +554,9 @@ void RasterizerOpenGL::InvalidateRegion(VAddr addr, u64 size, VideoCommon::Cache
     }
 }
 
-bool RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
+bool RasterizerOpenGL::OnCPUWrite(PAddr p_addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    const DAddr addr = device_memory.GetAddressFromPAddr(p_addr);
     if (addr == 0 || size == 0) {
         return false;
     }
@@ -576,8 +577,9 @@ bool RasterizerOpenGL::OnCPUWrite(VAddr addr, u64 size) {
     return false;
 }
 
-void RasterizerOpenGL::OnCacheInvalidation(VAddr addr, u64 size) {
+void RasterizerOpenGL::OnCacheInvalidation(PAddr p_addr, u64 size) {
     MICROPROFILE_SCOPE(OpenGL_CacheManagement);
+    const DAddr addr = device_memory.GetAddressFromPAddr(p_addr);
     if (addr == 0 || size == 0) {
         return;
     }
@@ -596,7 +598,7 @@ void RasterizerOpenGL::InvalidateGPUCache() {
     gpu.InvalidateGPUCache();
 }
 
-void RasterizerOpenGL::UnmapMemory(VAddr addr, u64 size) {
+void RasterizerOpenGL::UnmapMemory(DAddr addr, u64 size) {
     {
         std::scoped_lock lock{texture_cache.mutex};
         texture_cache.UnmapMemory(addr, size);
@@ -635,7 +637,7 @@ void RasterizerOpenGL::ReleaseFences(bool force) {
     fence_manager.WaitPendingFences(force);
 }
 
-void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size,
+void RasterizerOpenGL::FlushAndInvalidateRegion(DAddr addr, u64 size,
                                                 VideoCommon::CacheType which) {
     if (Settings::IsGPULevelExtreme()) {
         FlushRegion(addr, size, which);
@@ -739,7 +741,7 @@ void RasterizerOpenGL::AccelerateInlineToMemory(GPUVAddr address, size_t copy_si
 }
 
 bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
-                                         VAddr framebuffer_addr, u32 pixel_stride) {
+                                         DAddr framebuffer_addr, u32 pixel_stride) {
     if (framebuffer_addr == 0) {
         return false;
     }

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -14,7 +14,6 @@
 #include "common/common_types.h"
 #include "video_core/control/channel_state_cache.h"
 #include "video_core/engines/maxwell_dma.h"
-#include "video_core/rasterizer_accelerated.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/blit_image.h"
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
@@ -72,11 +71,11 @@ private:
     TextureCache& texture_cache;
 };
 
-class RasterizerOpenGL : public VideoCore::RasterizerAccelerated,
+class RasterizerOpenGL : public VideoCore::RasterizerInterface,
                          protected VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo> {
 public:
     explicit RasterizerOpenGL(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
-                              Core::Memory::Memory& cpu_memory_, const Device& device_,
+                              Tegra::MaxwellDeviceMemoryManager& device_memory_, const Device& device_,
                               ScreenInfo& screen_info_, ProgramManager& program_manager_,
                               StateTracker& state_tracker_);
     ~RasterizerOpenGL() override;
@@ -92,17 +91,17 @@ public:
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
     void DisableGraphicsUniformBuffer(size_t stage, u32 index) override;
     void FlushAll() override;
-    void FlushRegion(VAddr addr, u64 size,
+    void FlushRegion(DAddr addr, u64 size,
                      VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    bool MustFlushRegion(VAddr addr, u64 size,
+    bool MustFlushRegion(DAddr addr, u64 size,
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(PAddr addr, u64 size) override;
-    void InvalidateRegion(VAddr addr, u64 size,
+    void InvalidateRegion(DAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    void OnCacheInvalidation(VAddr addr, u64 size) override;
+    void OnCacheInvalidation(PAddr addr, u64 size) override;
-    bool OnCPUWrite(VAddr addr, u64 size) override;
+    bool OnCPUWrite(PAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
-    void UnmapMemory(VAddr addr, u64 size) override;
+    void UnmapMemory(DAddr addr, u64 size) override;
     void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) override;
     void SignalFence(std::function<void()>&& func) override;
     void SyncOperation(std::function<void()>&& func) override;
@@ -110,7 +109,7 @@ public:
     void SignalReference() override;
     void ReleaseFences(bool force = true) override;
     void FlushAndInvalidateRegion(
-        VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+        DAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void WaitForIdle() override;
     void FragmentBarrier() override;
     void TiledCacheBarrier() override;
@@ -123,7 +122,7 @@ public:
     Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() override;
     void AccelerateInlineToMemory(GPUVAddr address, size_t copy_size,
                                   std::span<const u8> memory) override;
-    bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
+    bool AccelerateDisplay(const Tegra::FramebufferConfig& config, DAddr framebuffer_addr,
                            u32 pixel_stride) override;
     void LoadDiskResources(u64 title_id, std::stop_token stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;
@@ -235,6 +234,7 @@ private:
                        VideoCommon::QueryPropertiesFlags flags, u32 payload, u32 subreport);
 
     Tegra::GPU& gpu;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
 
     const Device& device;
     ScreenInfo& screen_info;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -168,11 +168,12 @@ void SetXfbState(VideoCommon::TransformFeedbackState& state, const Maxwell& regs
 }
 } // Anonymous namespace
 
-ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindow& emu_window_,
+ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
-                         const Device& device_, TextureCache& texture_cache_,
+                         Core::Frontend::EmuWindow& emu_window_, const Device& device_,
-                         BufferCache& buffer_cache_, ProgramManager& program_manager_,
+                         TextureCache& texture_cache_, BufferCache& buffer_cache_,
-                         StateTracker& state_tracker_, VideoCore::ShaderNotify& shader_notify_)
+                         ProgramManager& program_manager_, StateTracker& state_tracker_,
-    : VideoCommon::ShaderCache{rasterizer_}, emu_window{emu_window_}, device{device_},
+                         VideoCore::ShaderNotify& shader_notify_)
+    : VideoCommon::ShaderCache{device_memory_}, emu_window{emu_window_}, device{device_},
       texture_cache{texture_cache_}, buffer_cache{buffer_cache_}, program_manager{program_manager_},
       state_tracker{state_tracker_}, shader_notify{shader_notify_},
       use_asynchronous_shaders{device.UseAsynchronousShaders()},

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -17,7 +17,7 @@
 
 namespace Tegra {
 class MemoryManager;
-}
+} // namespace Tegra
 
 namespace OpenGL {
 
@@ -28,10 +28,11 @@ using ShaderWorker = Common::StatefulThreadWorker<ShaderContext::Context>;
 
 class ShaderCache : public VideoCommon::ShaderCache {
 public:
-    explicit ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindow& emu_window_,
+    explicit ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_,
-                         const Device& device_, TextureCache& texture_cache_,
+                         Core::Frontend::EmuWindow& emu_window_, const Device& device_,
-                         BufferCache& buffer_cache_, ProgramManager& program_manager_,
+                         TextureCache& texture_cache_, BufferCache& buffer_cache_,
-                         StateTracker& state_tracker_, VideoCore::ShaderNotify& shader_notify_);
+                         ProgramManager& program_manager_, StateTracker& state_tracker_,
+                         VideoCore::ShaderNotify& shader_notify_);
     ~ShaderCache();
 
     void LoadDiskResources(u64 title_id, std::stop_token stop_loading,

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -144,12 +144,13 @@ void APIENTRY DebugHandler(GLenum source, GLenum type, GLuint id, GLenum severit
 
 RendererOpenGL::RendererOpenGL(Core::TelemetrySession& telemetry_session_,
                                Core::Frontend::EmuWindow& emu_window_,
-                               Core::Memory::Memory& cpu_memory_, Tegra::GPU& gpu_,
+                               Tegra::MaxwellDeviceMemoryManager& device_memory_, Tegra::GPU& gpu_,
                                std::unique_ptr<Core::Frontend::GraphicsContext> context_)
     : RendererBase{emu_window_, std::move(context_)}, telemetry_session{telemetry_session_},
-      emu_window{emu_window_}, cpu_memory{cpu_memory_}, gpu{gpu_}, device{emu_window_},
+      emu_window{emu_window_}, device_memory{device_memory_}, gpu{gpu_}, device{emu_window_},
       state_tracker{}, program_manager{device},
-      rasterizer(emu_window, gpu, cpu_memory, device, screen_info, program_manager, state_tracker) {
+      rasterizer(emu_window, gpu, device_memory, device, screen_info, program_manager,
+                 state_tracker) {
     if (Settings::values.renderer_debug && GLAD_GL_KHR_debug) {
         glEnable(GL_DEBUG_OUTPUT);
         glEnable(GL_DEBUG_OUTPUT_SYNCHRONOUS);
@@ -242,7 +243,7 @@ void RendererOpenGL::LoadFBToScreenInfo(const Tegra::FramebufferConfig& framebuf
     const u32 bytes_per_pixel{VideoCore::Surface::BytesPerBlock(pixel_format)};
     const u64 size_in_bytes{Tegra::Texture::CalculateSize(
         true, bytes_per_pixel, framebuffer.stride, framebuffer.height, 1, block_height_log2, 0)};
-    const u8* const host_ptr{cpu_memory.GetPointer(framebuffer_addr)};
+    const u8* const host_ptr{device_memory.GetPointer<u8>(framebuffer_addr)};
     const std::span<const u8> input_data(host_ptr, size_in_bytes);
     Tegra::Texture::UnswizzleTexture(gl_framebuffer_data, input_data, bytes_per_pixel,
                                      framebuffer.width, framebuffer.height, 1, block_height_log2,

src/video_core/renderer_opengl/renderer_opengl.h

@@ -61,7 +61,7 @@ class RendererOpenGL final : public VideoCore::RendererBase {
 public:
     explicit RendererOpenGL(Core::TelemetrySession& telemetry_session_,
                             Core::Frontend::EmuWindow& emu_window_,
-                            Core::Memory::Memory& cpu_memory_, Tegra::GPU& gpu_,
+                            Tegra::MaxwellDeviceMemoryManager& device_memory_, Tegra::GPU& gpu_,
                             std::unique_ptr<Core::Frontend::GraphicsContext> context_);
     ~RendererOpenGL() override;
 
@@ -101,7 +101,7 @@ private:
 
     Core::TelemetrySession& telemetry_session;
     Core::Frontend::EmuWindow& emu_window;
-    Core::Memory::Memory& cpu_memory;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
     Tegra::GPU& gpu;
 
     Device device;

src/video_core/renderer_vulkan/renderer_vulkan.cpp

@@ -82,10 +82,10 @@ Device CreateDevice(const vk::Instance& instance, const vk::InstanceDispatch& dl
 
 RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
                                Core::Frontend::EmuWindow& emu_window,
-                               Core::Memory::Memory& cpu_memory_, Tegra::GPU& gpu_,
+                               Tegra::MaxwellDeviceMemoryManager& device_memory_, Tegra::GPU& gpu_,
                                std::unique_ptr<Core::Frontend::GraphicsContext> context_) try
     : RendererBase(emu_window, std::move(context_)), telemetry_session(telemetry_session_),
-      cpu_memory(cpu_memory_), gpu(gpu_), library(OpenLibrary(context.get())),
+      device_memory(device_memory_), gpu(gpu_), library(OpenLibrary(context.get())),
       instance(CreateInstance(*library, dld, VK_API_VERSION_1_1, render_window.GetWindowInfo().type,
                               Settings::values.renderer_debug.GetValue())),
       debug_messenger(Settings::values.renderer_debug ? CreateDebugUtilsCallback(instance)
@@ -97,9 +97,9 @@ RendererVulkan::RendererVulkan(Core::TelemetrySession& telemetry_session_,
                 render_window.GetFramebufferLayout().height),
       present_manager(instance, render_window, device, memory_allocator, scheduler, swapchain,
                       surface),
-      blit_screen(cpu_memory, render_window, device, memory_allocator, swapchain, present_manager,
+      blit_screen(device_memory, render_window, device, memory_allocator, swapchain, present_manager,
                   scheduler, screen_info),
-      rasterizer(render_window, gpu, cpu_memory, screen_info, device, memory_allocator,
+      rasterizer(render_window, gpu, device_memory, screen_info, device, memory_allocator,
                  state_tracker, scheduler) {
     if (Settings::values.renderer_force_max_clock.GetValue() && device.ShouldBoostClocks()) {
         turbo_mode.emplace(instance, dld);
@@ -128,7 +128,7 @@ void RendererVulkan::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) {
     screen_info.width = framebuffer->width;
     screen_info.height = framebuffer->height;
 
-    const VAddr framebuffer_addr = framebuffer->address + framebuffer->offset;
+    const DAddr framebuffer_addr = framebuffer->address + framebuffer->offset;
     const bool use_accelerated =
         rasterizer.AccelerateDisplay(*framebuffer, framebuffer_addr, framebuffer->stride);
     RenderScreenshot(*framebuffer, use_accelerated);

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -20,6 +20,7 @@
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 
 namespace Core {
 class TelemetrySession;
@@ -42,7 +43,7 @@ class RendererVulkan final : public VideoCore::RendererBase {
 public:
     explicit RendererVulkan(Core::TelemetrySession& telemtry_session,
                             Core::Frontend::EmuWindow& emu_window,
-                            Core::Memory::Memory& cpu_memory_, Tegra::GPU& gpu_,
+                            Tegra::MaxwellDeviceMemoryManager& device_memory_, Tegra::GPU& gpu_,
                             std::unique_ptr<Core::Frontend::GraphicsContext> context_);
     ~RendererVulkan() override;
 
@@ -62,7 +63,7 @@ private:
     void RenderScreenshot(const Tegra::FramebufferConfig& framebuffer, bool use_accelerated);
 
     Core::TelemetrySession& telemetry_session;
-    Core::Memory::Memory& cpu_memory;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
     Tegra::GPU& gpu;
 
     std::shared_ptr<Common::DynamicLibrary> library;

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -14,8 +14,8 @@
 #include "common/settings.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
-#include "core/memory.h"
 #include "video_core/gpu.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/host_shaders/fxaa_frag_spv.h"
 #include "video_core/host_shaders/fxaa_vert_spv.h"
 #include "video_core/host_shaders/present_bicubic_frag_spv.h"
@@ -121,11 +121,12 @@ struct BlitScreen::BufferData {
     // Unaligned image data goes here
 };
 
-BlitScreen::BlitScreen(Core::Memory::Memory& cpu_memory_, Core::Frontend::EmuWindow& render_window_,
+BlitScreen::BlitScreen(Tegra::MaxwellDeviceMemoryManager& device_memory_,
-                       const Device& device_, MemoryAllocator& memory_allocator_,
+                       Core::Frontend::EmuWindow& render_window_, const Device& device_,
-                       Swapchain& swapchain_, PresentManager& present_manager_,
+                       MemoryAllocator& memory_allocator_, Swapchain& swapchain_,
-                       Scheduler& scheduler_, const ScreenInfo& screen_info_)
+                       PresentManager& present_manager_, Scheduler& scheduler_,
-    : cpu_memory{cpu_memory_}, render_window{render_window_}, device{device_},
+                       const ScreenInfo& screen_info_)
+    : device_memory{device_memory_}, render_window{render_window_}, device{device_},
       memory_allocator{memory_allocator_}, swapchain{swapchain_}, present_manager{present_manager_},
       scheduler{scheduler_}, image_count{swapchain.GetImageCount()}, screen_info{screen_info_} {
     resource_ticks.resize(image_count);
@@ -219,8 +220,8 @@ void BlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
     if (!use_accelerated) {
         const u64 image_offset = GetRawImageOffset(framebuffer);
 
-        const VAddr framebuffer_addr = framebuffer.address + framebuffer.offset;
+        const DAddr framebuffer_addr = framebuffer.address + framebuffer.offset;
-        const u8* const host_ptr = cpu_memory.GetPointer(framebuffer_addr);
+        const u8* const host_ptr = device_memory.GetPointer<u8>(framebuffer_addr);
 
         // TODO(Rodrigo): Read this from HLE
         constexpr u32 block_height_log2 = 4;

src/video_core/renderer_vulkan/vk_blit_screen.h

@@ -8,15 +8,12 @@
 #include "core/frontend/framebuffer_layout.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 
 namespace Core {
 class System;
 }
 
-namespace Core::Memory {
-class Memory;
-}
-
 namespace Core::Frontend {
 class EmuWindow;
 }
@@ -56,7 +53,7 @@ struct ScreenInfo {
 
 class BlitScreen {
 public:
-    explicit BlitScreen(Core::Memory::Memory& cpu_memory, Core::Frontend::EmuWindow& render_window,
+    explicit BlitScreen(Tegra::MaxwellDeviceMemoryManager& device_memory, Core::Frontend::EmuWindow& render_window,
                         const Device& device, MemoryAllocator& memory_manager, Swapchain& swapchain,
                         PresentManager& present_manager, Scheduler& scheduler,
                         const ScreenInfo& screen_info);
@@ -109,7 +106,7 @@ private:
     u64 CalculateBufferSize(const Tegra::FramebufferConfig& framebuffer) const;
     u64 GetRawImageOffset(const Tegra::FramebufferConfig& framebuffer) const;
 
-    Core::Memory::Memory& cpu_memory;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
     Core::Frontend::EmuWindow& render_window;
     const Device& device;
     MemoryAllocator& memory_allocator;

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -79,7 +79,7 @@ vk::Buffer CreateBuffer(const Device& device, const MemoryAllocator& memory_allo
 } // Anonymous namespace
 
 Buffer::Buffer(BufferCacheRuntime& runtime, VideoCommon::NullBufferParams null_params)
-    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(null_params), tracker{4096} {
+    : VideoCommon::BufferBase(null_params), tracker{4096} {
     if (runtime.device.HasNullDescriptor()) {
         return;
     }
@@ -88,11 +88,9 @@ Buffer::Buffer(BufferCacheRuntime& runtime, VideoCommon::NullBufferParams null_p
     is_null = true;
 }
 
-Buffer::Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
+Buffer::Buffer(BufferCacheRuntime& runtime, DAddr cpu_addr_, u64 size_bytes_)
-               VAddr cpu_addr_, u64 size_bytes_)
+    : VideoCommon::BufferBase(cpu_addr_, size_bytes_), device{&runtime.device},
-    : VideoCommon::BufferBase<VideoCore::RasterizerInterface>(rasterizer_, cpu_addr_, size_bytes_),
+      buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes())}, tracker{SizeBytes()} {
-      device{&runtime.device}, buffer{CreateBuffer(*device, runtime.memory_allocator, SizeBytes())},
-      tracker{SizeBytes()} {
     if (runtime.device.HasDebuggingToolAttached()) {
         buffer.SetObjectNameEXT(fmt::format("Buffer 0x{:x}", CpuAddr()).c_str());
     }

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -23,11 +23,10 @@ struct HostVertexBinding;
 
 class BufferCacheRuntime;
 
-class Buffer : public VideoCommon::BufferBase<VideoCore::RasterizerInterface> {
+class Buffer : public VideoCommon::BufferBase {
 public:
     explicit Buffer(BufferCacheRuntime&, VideoCommon::NullBufferParams null_params);
-    explicit Buffer(BufferCacheRuntime& runtime, VideoCore::RasterizerInterface& rasterizer_,
+    explicit Buffer(BufferCacheRuntime& runtime, VAddr cpu_addr_, u64 size_bytes_);
-                    VAddr cpu_addr_, u64 size_bytes_);
 
     [[nodiscard]] VkBufferView View(u32 offset, u32 size, VideoCore::Surface::PixelFormat format);
 
@@ -173,7 +172,7 @@ struct BufferCacheParams {
     using Runtime = Vulkan::BufferCacheRuntime;
     using Buffer = Vulkan::Buffer;
     using Async_Buffer = Vulkan::StagingBufferRef;
-    using MemoryTracker = VideoCommon::MemoryTrackerBase<VideoCore::RasterizerInterface>;
+    using MemoryTracker = VideoCommon::MemoryTrackerBase<Tegra::MaxwellDeviceMemoryManager>;
 
     static constexpr bool IS_OPENGL = false;
     static constexpr bool HAS_PERSISTENT_UNIFORM_BUFFER_BINDINGS = false;

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -30,7 +30,6 @@
 #include "video_core/renderer_vulkan/vk_compute_pipeline.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
-#include "video_core/renderer_vulkan/vk_rasterizer.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_shader_util.h"
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
@@ -299,12 +298,12 @@ bool GraphicsPipelineCacheKey::operator==(const GraphicsPipelineCacheKey& rhs) c
     return std::memcmp(&rhs, this, Size()) == 0;
 }
 
-PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device_,
+PipelineCache::PipelineCache(Tegra::MaxwellDeviceMemoryManager& device_memory_, const Device& device_,
                              Scheduler& scheduler_, DescriptorPool& descriptor_pool_,
                              GuestDescriptorQueue& guest_descriptor_queue_,
                              RenderPassCache& render_pass_cache_, BufferCache& buffer_cache_,
                              TextureCache& texture_cache_, VideoCore::ShaderNotify& shader_notify_)
-    : VideoCommon::ShaderCache{rasterizer_}, device{device_}, scheduler{scheduler_},
+    : VideoCommon::ShaderCache{device_memory_}, device{device_}, scheduler{scheduler_},
       descriptor_pool{descriptor_pool_}, guest_descriptor_queue{guest_descriptor_queue_},
       render_pass_cache{render_pass_cache_}, buffer_cache{buffer_cache_},
       texture_cache{texture_cache_}, shader_notify{shader_notify_},

src/video_core/renderer_vulkan/vk_pipeline_cache.h

@@ -26,6 +26,7 @@
 #include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
 #include "video_core/renderer_vulkan/vk_texture_cache.h"
 #include "video_core/shader_cache.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 
 namespace Core {
 class System;
@@ -79,7 +80,6 @@ class ComputePipeline;
 class DescriptorPool;
 class Device;
 class PipelineStatistics;
-class RasterizerVulkan;
 class RenderPassCache;
 class Scheduler;
 
@@ -99,7 +99,7 @@ struct ShaderPools {
 
 class PipelineCache : public VideoCommon::ShaderCache {
 public:
-    explicit PipelineCache(RasterizerVulkan& rasterizer, const Device& device, Scheduler& scheduler,
+    explicit PipelineCache(Tegra::MaxwellDeviceMemoryManager& device_memory_, const Device& device, Scheduler& scheduler,
                            DescriptorPool& descriptor_pool,
                            GuestDescriptorQueue& guest_descriptor_queue,
                            RenderPassCache& render_pass_cache, BufferCache& buffer_cache,

src/video_core/renderer_vulkan/vk_query_cache.cpp

@@ -14,7 +14,9 @@
 #include "common/bit_util.h"
 #include "common/common_types.h"
 #include "core/memory.h"
+#include "video_core/rasterizer_interface.h"
 #include "video_core/engines/draw_manager.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/query_cache/query_cache.h"
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
 #include "video_core/renderer_vulkan/vk_compute_pass.h"
@@ -102,7 +104,7 @@ private:
 using BaseStreamer = VideoCommon::SimpleStreamer<VideoCommon::HostQueryBase>;
 
 struct HostSyncValues {
-    VAddr address;
+    DAddr address;
     size_t size;
     size_t offset;
 
@@ -317,7 +319,7 @@ public:
         pending_sync.clear();
     }
 
-    size_t WriteCounter(VAddr address, bool has_timestamp, u32 value,
+    size_t WriteCounter(DAddr address, bool has_timestamp, u32 value,
                         [[maybe_unused]] std::optional<u32> subreport) override {
         PauseCounter();
         auto index = BuildQuery();
@@ -738,7 +740,7 @@ public:
         pending_sync.clear();
     }
 
-    size_t WriteCounter(VAddr address, bool has_timestamp, u32 value,
+    size_t WriteCounter(DAddr address, bool has_timestamp, u32 value,
                         std::optional<u32> subreport_) override {
         auto index = BuildQuery();
         auto* new_query = GetQuery(index);
@@ -769,9 +771,9 @@ public:
         return index;
     }
 
-    std::optional<std::pair<VAddr, size_t>> GetLastQueryStream(size_t stream) {
+    std::optional<std::pair<DAddr, size_t>> GetLastQueryStream(size_t stream) {
         if (last_queries[stream] != 0) {
-            std::pair<VAddr, size_t> result(last_queries[stream], last_queries_stride[stream]);
+            std::pair<DAddr, size_t> result(last_queries[stream], last_queries_stride[stream]);
             return result;
         }
         return std::nullopt;
@@ -974,7 +976,7 @@ private:
     size_t buffers_count{};
     std::array<VkBuffer, NUM_STREAMS> counter_buffers{};
     std::array<VkDeviceSize, NUM_STREAMS> offsets{};
-    std::array<VAddr, NUM_STREAMS> last_queries;
+    std::array<DAddr, NUM_STREAMS> last_queries;
     std::array<size_t, NUM_STREAMS> last_queries_stride;
     Maxwell3D::Regs::PrimitiveTopology out_topology;
     u64 streams_mask;
@@ -987,7 +989,7 @@ public:
         : VideoCommon::QueryBase(0, VideoCommon::QueryFlagBits::IsHostManaged, 0) {}
 
     // Parameterized constructor
-    PrimitivesQueryBase(bool has_timestamp, VAddr address)
+    PrimitivesQueryBase(bool has_timestamp, DAddr address)
         : VideoCommon::QueryBase(address, VideoCommon::QueryFlagBits::IsHostManaged, 0) {
         if (has_timestamp) {
             flags |= VideoCommon::QueryFlagBits::HasTimestamp;
@@ -995,7 +997,7 @@ public:
     }
 
     u64 stride{};
-    VAddr dependant_address{};
+    DAddr dependant_address{};
     Maxwell3D::Regs::PrimitiveTopology topology{Maxwell3D::Regs::PrimitiveTopology::Points};
     size_t dependant_index{};
     bool dependant_manage{};
@@ -1005,15 +1007,15 @@ class PrimitivesSucceededStreamer : public VideoCommon::SimpleStreamer<Primitive
 public:
     explicit PrimitivesSucceededStreamer(size_t id_, QueryCacheRuntime& runtime_,
                                          TFBCounterStreamer& tfb_streamer_,
-                                         Core::Memory::Memory& cpu_memory_)
+                                         Tegra::MaxwellDeviceMemoryManager& device_memory_)
         : VideoCommon::SimpleStreamer<PrimitivesQueryBase>(id_), runtime{runtime_},
-          tfb_streamer{tfb_streamer_}, cpu_memory{cpu_memory_} {
+          tfb_streamer{tfb_streamer_}, device_memory{device_memory_} {
         MakeDependent(&tfb_streamer);
     }
 
     ~PrimitivesSucceededStreamer() = default;
 
-    size_t WriteCounter(VAddr address, bool has_timestamp, u32 value,
+    size_t WriteCounter(DAddr address, bool has_timestamp, u32 value,
                         std::optional<u32> subreport_) override {
         auto index = BuildQuery();
         auto* new_query = GetQuery(index);
@@ -1063,6 +1065,8 @@ public:
                 }
             });
         }
+        auto* ptr = device_memory.GetPointer<u8>(new_query->dependant_address);
+        ASSERT(ptr != nullptr);
 
         new_query->dependant_manage = must_manage_dependance;
         pending_flush_queries.push_back(index);
@@ -1100,7 +1104,7 @@ public:
                 num_vertices = dependant_query->value / query->stride;
                 tfb_streamer.Free(query->dependant_index);
             } else {
-                u8* pointer = cpu_memory.GetPointer(query->dependant_address);
+                u8* pointer = device_memory.GetPointer<u8>(query->dependant_address);
                 u32 result;
                 std::memcpy(&result, pointer, sizeof(u32));
                 num_vertices = static_cast<u64>(result) / query->stride;
@@ -1137,7 +1141,7 @@ public:
 private:
     QueryCacheRuntime& runtime;
     TFBCounterStreamer& tfb_streamer;
-    Core::Memory::Memory& cpu_memory;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
 
     // syncing queue
     std::vector<size_t> pending_sync;
@@ -1152,12 +1156,12 @@ private:
 
 struct QueryCacheRuntimeImpl {
     QueryCacheRuntimeImpl(QueryCacheRuntime& runtime, VideoCore::RasterizerInterface* rasterizer_,
-                          Core::Memory::Memory& cpu_memory_, Vulkan::BufferCache& buffer_cache_,
+                          Tegra::MaxwellDeviceMemoryManager& device_memory_, Vulkan::BufferCache& buffer_cache_,
                           const Device& device_, const MemoryAllocator& memory_allocator_,
                           Scheduler& scheduler_, StagingBufferPool& staging_pool_,
                           ComputePassDescriptorQueue& compute_pass_descriptor_queue,
                           DescriptorPool& descriptor_pool)
-        : rasterizer{rasterizer_}, cpu_memory{cpu_memory_},
+        : rasterizer{rasterizer_}, device_memory{device_memory_},
           buffer_cache{buffer_cache_}, device{device_},
           memory_allocator{memory_allocator_}, scheduler{scheduler_}, staging_pool{staging_pool_},
           guest_streamer(0, runtime),
@@ -1168,7 +1172,7 @@ struct QueryCacheRuntimeImpl {
                        scheduler, memory_allocator, staging_pool),
           primitives_succeeded_streamer(
               static_cast<size_t>(QueryType::StreamingPrimitivesSucceeded), runtime, tfb_streamer,
-              cpu_memory_),
+              device_memory_),
           primitives_needed_minus_succeeded_streamer(
               static_cast<size_t>(QueryType::StreamingPrimitivesNeededMinusSucceeded), runtime, 0u),
           hcr_setup{}, hcr_is_set{}, is_hcr_running{}, maxwell3d{} {
@@ -1195,7 +1199,7 @@ struct QueryCacheRuntimeImpl {
     }
 
     VideoCore::RasterizerInterface* rasterizer;
-    Core::Memory::Memory& cpu_memory;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
     Vulkan::BufferCache& buffer_cache;
 
     const Device& device;
@@ -1210,7 +1214,7 @@ struct QueryCacheRuntimeImpl {
     PrimitivesSucceededStreamer primitives_succeeded_streamer;
     VideoCommon::StubStreamer<QueryCacheParams> primitives_needed_minus_succeeded_streamer;
 
-    std::vector<std::pair<VAddr, VAddr>> little_cache;
+    std::vector<std::pair<DAddr, DAddr>> little_cache;
     std::vector<std::pair<VkBuffer, VkDeviceSize>> buffers_to_upload_to;
     std::vector<size_t> redirect_cache;
     std::vector<std::vector<VkBufferCopy>> copies_setup;
@@ -1229,14 +1233,14 @@ struct QueryCacheRuntimeImpl {
 };
 
 QueryCacheRuntime::QueryCacheRuntime(VideoCore::RasterizerInterface* rasterizer,
-                                     Core::Memory::Memory& cpu_memory_,
+                                     Tegra::MaxwellDeviceMemoryManager& device_memory_,
                                      Vulkan::BufferCache& buffer_cache_, const Device& device_,
                                      const MemoryAllocator& memory_allocator_,
                                      Scheduler& scheduler_, StagingBufferPool& staging_pool_,
                                      ComputePassDescriptorQueue& compute_pass_descriptor_queue,
                                      DescriptorPool& descriptor_pool) {
     impl = std::make_unique<QueryCacheRuntimeImpl>(
-        *this, rasterizer, cpu_memory_, buffer_cache_, device_, memory_allocator_, scheduler_,
+        *this, rasterizer, device_memory_, buffer_cache_, device_, memory_allocator_, scheduler_,
         staging_pool_, compute_pass_descriptor_queue, descriptor_pool);
 }
 
@@ -1309,7 +1313,7 @@ void QueryCacheRuntime::HostConditionalRenderingCompareValueImpl(VideoCommon::Lo
     ResumeHostConditionalRendering();
 }
 
-void QueryCacheRuntime::HostConditionalRenderingCompareBCImpl(VAddr address, bool is_equal) {
+void QueryCacheRuntime::HostConditionalRenderingCompareBCImpl(DAddr address, bool is_equal) {
     VkBuffer to_resolve;
     u32 to_resolve_offset;
     {
@@ -1350,11 +1354,11 @@ bool QueryCacheRuntime::HostConditionalRenderingCompareValues(VideoCommon::Looku
         return false;
     }
 
-    const auto check_in_bc = [&](VAddr address) {
+    const auto check_in_bc = [&](DAddr address) {
         return impl->buffer_cache.IsRegionGpuModified(address, 8);
     };
-    const auto check_value = [&](VAddr address) {
+    const auto check_value = [&](DAddr address) {
-        u8* ptr = impl->cpu_memory.GetPointer(address);
+        u8* ptr = impl->device_memory.GetPointer<u8>(address);
         u64 value{};
         std::memcpy(&value, ptr, sizeof(value));
         return value == 0;
@@ -1477,8 +1481,8 @@ void QueryCacheRuntime::SyncValues(std::span<SyncValuesType> values, VkBuffer ba
     for (auto& sync_val : values) {
         total_size += sync_val.size;
         bool found = false;
-        VAddr base = Common::AlignDown(sync_val.address, Core::Memory::YUZU_PAGESIZE);
+        DAddr base = Common::AlignDown(sync_val.address, Core::Memory::YUZU_PAGESIZE);
-        VAddr base_end = base + Core::Memory::YUZU_PAGESIZE;
+        DAddr base_end = base + Core::Memory::YUZU_PAGESIZE;
         for (size_t i = 0; i < impl->little_cache.size(); i++) {
             const auto set_found = [&] {
                 impl->redirect_cache.push_back(i);

src/video_core/renderer_vulkan/vk_query_cache.h

@@ -27,7 +27,7 @@ struct QueryCacheRuntimeImpl;
 class QueryCacheRuntime {
 public:
     explicit QueryCacheRuntime(VideoCore::RasterizerInterface* rasterizer,
-                               Core::Memory::Memory& cpu_memory_,
+                               Tegra::MaxwellDeviceMemoryManager& device_memory_,
                                Vulkan::BufferCache& buffer_cache_, const Device& device_,
                                const MemoryAllocator& memory_allocator_, Scheduler& scheduler_,
                                StagingBufferPool& staging_pool_,
@@ -61,7 +61,7 @@ public:
 
 private:
     void HostConditionalRenderingCompareValueImpl(VideoCommon::LookupData object, bool is_equal);
-    void HostConditionalRenderingCompareBCImpl(VAddr address, bool is_equal);
+    void HostConditionalRenderingCompareBCImpl(DAddr address, bool is_equal);
     friend struct QueryCacheRuntimeImpl;
     std::unique_ptr<QueryCacheRuntimeImpl> impl;
 };

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -18,6 +18,7 @@
 #include "video_core/engines/draw_manager.h"
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/renderer_vulkan/blit_image.h"
 #include "video_core/renderer_vulkan/fixed_pipeline_state.h"
 #include "video_core/renderer_vulkan/maxwell_to_vk.h"
@@ -37,6 +38,7 @@
 #include "video_core/vulkan_common/vulkan_device.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
 
+
 namespace Vulkan {
 
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
@@ -163,10 +165,11 @@ DrawParams MakeDrawParams(const MaxwellDrawState& draw_state, u32 num_instances,
 } // Anonymous namespace
 
 RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
-                                   Core::Memory::Memory& cpu_memory_, ScreenInfo& screen_info_,
+                                   Tegra::MaxwellDeviceMemoryManager& device_memory_,
-                                   const Device& device_, MemoryAllocator& memory_allocator_,
+                                   ScreenInfo& screen_info_, const Device& device_,
-                                   StateTracker& state_tracker_, Scheduler& scheduler_)
+                                   MemoryAllocator& memory_allocator_, StateTracker& state_tracker_,
-    : RasterizerAccelerated{cpu_memory_}, gpu{gpu_}, screen_info{screen_info_}, device{device_},
+                                   Scheduler& scheduler_)
+    : gpu{gpu_}, device_memory{device_memory_}, screen_info{screen_info_}, device{device_},
       memory_allocator{memory_allocator_}, state_tracker{state_tracker_}, scheduler{scheduler_},
       staging_pool(device, memory_allocator, scheduler), descriptor_pool(device, scheduler),
       guest_descriptor_queue(device, scheduler), compute_pass_descriptor_queue(device, scheduler),
@@ -174,14 +177,14 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
       texture_cache_runtime{
           device,     scheduler,         memory_allocator, staging_pool,
           blit_image, render_pass_cache, descriptor_pool,  compute_pass_descriptor_queue},
-      texture_cache(texture_cache_runtime, *this),
+      texture_cache(texture_cache_runtime, device_memory),
       buffer_cache_runtime(device, memory_allocator, scheduler, staging_pool,
                            guest_descriptor_queue, compute_pass_descriptor_queue, descriptor_pool),
-      buffer_cache(*this, cpu_memory_, buffer_cache_runtime),
+      buffer_cache(device_memory, buffer_cache_runtime),
-      query_cache_runtime(this, cpu_memory_, buffer_cache, device, memory_allocator, scheduler,
+      query_cache_runtime(this, device_memory, buffer_cache, device, memory_allocator, scheduler,
                           staging_pool, compute_pass_descriptor_queue, descriptor_pool),
-      query_cache(gpu, *this, cpu_memory_, query_cache_runtime),
+      query_cache(gpu, *this, device_memory, query_cache_runtime),
-      pipeline_cache(*this, device, scheduler, descriptor_pool, guest_descriptor_queue,
+      pipeline_cache(device_memory, device, scheduler, descriptor_pool, guest_descriptor_queue,
                      render_pass_cache, buffer_cache, texture_cache, gpu.ShaderNotify()),
       accelerate_dma(buffer_cache, texture_cache, scheduler),
       fence_manager(*this, gpu, texture_cache, buffer_cache, query_cache, device, scheduler),
@@ -508,7 +511,7 @@ void Vulkan::RasterizerVulkan::DisableGraphicsUniformBuffer(size_t stage, u32 in
 
 void RasterizerVulkan::FlushAll() {}
 
-void RasterizerVulkan::FlushRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
+void RasterizerVulkan::FlushRegion(DAddr addr, u64 size, VideoCommon::CacheType which) {
     if (addr == 0 || size == 0) {
         return;
     }
@@ -525,7 +528,7 @@ void RasterizerVulkan::FlushRegion(VAddr addr, u64 size, VideoCommon::CacheType
     }
 }
 
-bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
+bool RasterizerVulkan::MustFlushRegion(DAddr addr, u64 size, VideoCommon::CacheType which) {
     if ((True(which & VideoCommon::CacheType::BufferCache))) {
         std::scoped_lock lock{buffer_cache.mutex};
         if (buffer_cache.IsRegionGpuModified(addr, size)) {
@@ -542,7 +545,7 @@ bool RasterizerVulkan::MustFlushRegion(VAddr addr, u64 size, VideoCommon::CacheT
     return false;
 }
 
-VideoCore::RasterizerDownloadArea RasterizerVulkan::GetFlushArea(VAddr addr, u64 size) {
+VideoCore::RasterizerDownloadArea RasterizerVulkan::GetFlushArea(DAddr addr, u64 size) {
     {
         std::scoped_lock lock{texture_cache.mutex};
         auto area = texture_cache.GetFlushArea(addr, size);
@@ -558,7 +561,7 @@ VideoCore::RasterizerDownloadArea RasterizerVulkan::GetFlushArea(VAddr addr, u64
     return new_area;
 }
 
-void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size, VideoCommon::CacheType which) {
+void RasterizerVulkan::InvalidateRegion(DAddr addr, u64 size, VideoCommon::CacheType which) {
     if (addr == 0 || size == 0) {
         return;
     }
@@ -578,7 +581,7 @@ void RasterizerVulkan::InvalidateRegion(VAddr addr, u64 size, VideoCommon::Cache
     }
 }
 
-void RasterizerVulkan::InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) {
+void RasterizerVulkan::InnerInvalidation(std::span<const std::pair<DAddr, std::size_t>> sequences) {
     {
         std::scoped_lock lock{texture_cache.mutex};
         for (const auto& [addr, size] : sequences) {
@@ -599,7 +602,8 @@ void RasterizerVulkan::InnerInvalidation(std::span<const std::pair<VAddr, std::s
     }
 }
 
-bool RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
+bool RasterizerVulkan::OnCPUWrite(PAddr p_addr, u64 size) {
+    const DAddr addr = device_memory.GetAddressFromPAddr(p_addr);
     if (addr == 0 || size == 0) {
         return false;
     }
@@ -620,7 +624,8 @@ bool RasterizerVulkan::OnCPUWrite(VAddr addr, u64 size) {
     return false;
 }
 
-void RasterizerVulkan::OnCacheInvalidation(VAddr addr, u64 size) {
+void RasterizerVulkan::OnCacheInvalidation(PAddr p_addr, u64 size) {
+    const DAddr addr = device_memory.GetAddressFromPAddr(p_addr);
     if (addr == 0 || size == 0) {
         return;
     }
@@ -640,7 +645,7 @@ void RasterizerVulkan::InvalidateGPUCache() {
     gpu.InvalidateGPUCache();
 }
 
-void RasterizerVulkan::UnmapMemory(VAddr addr, u64 size) {
+void RasterizerVulkan::UnmapMemory(DAddr addr, u64 size) {
     {
         std::scoped_lock lock{texture_cache.mutex};
         texture_cache.UnmapMemory(addr, size);
@@ -679,7 +684,7 @@ void RasterizerVulkan::ReleaseFences(bool force) {
     fence_manager.WaitPendingFences(force);
 }
 
-void RasterizerVulkan::FlushAndInvalidateRegion(VAddr addr, u64 size,
+void RasterizerVulkan::FlushAndInvalidateRegion(DAddr addr, u64 size,
                                                 VideoCommon::CacheType which) {
     if (Settings::IsGPULevelExtreme()) {
         FlushRegion(addr, size, which);
@@ -782,7 +787,7 @@ void RasterizerVulkan::AccelerateInlineToMemory(GPUVAddr address, size_t copy_si
 }
 
 bool RasterizerVulkan::AccelerateDisplay(const Tegra::FramebufferConfig& config,
-                                         VAddr framebuffer_addr, u32 pixel_stride) {
+                                         DAddr framebuffer_addr, u32 pixel_stride) {
     if (!framebuffer_addr) {
         return false;
     }

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -12,7 +12,6 @@
 #include "common/common_types.h"
 #include "video_core/control/channel_state_cache.h"
 #include "video_core/engines/maxwell_dma.h"
-#include "video_core/rasterizer_accelerated.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_vulkan/blit_image.h"
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
@@ -25,6 +24,7 @@
 #include "video_core/renderer_vulkan/vk_update_descriptor.h"
 #include "video_core/vulkan_common/vulkan_memory_allocator.h"
 #include "video_core/vulkan_common/vulkan_wrapper.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 
 namespace Core {
 class System;
@@ -34,10 +34,14 @@ namespace Core::Frontend {
 class EmuWindow;
 }
 
-namespace Tegra::Engines {
+namespace Tegra {
+
+namespace Engines {
 class Maxwell3D;
 }
 
+} // namespace Tegra
+
 namespace Vulkan {
 
 struct ScreenInfo;
@@ -70,13 +74,14 @@ private:
     Scheduler& scheduler;
 };
 
-class RasterizerVulkan final : public VideoCore::RasterizerAccelerated,
+class RasterizerVulkan final : public VideoCore::RasterizerInterface,
                                protected VideoCommon::ChannelSetupCaches<VideoCommon::ChannelInfo> {
 public:
     explicit RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra::GPU& gpu_,
-                              Core::Memory::Memory& cpu_memory_, ScreenInfo& screen_info_,
+                              Tegra::MaxwellDeviceMemoryManager& device_memory_,
-                              const Device& device_, MemoryAllocator& memory_allocator_,
+                              ScreenInfo& screen_info_, const Device& device_,
-                              StateTracker& state_tracker_, Scheduler& scheduler_);
+                              MemoryAllocator& memory_allocator_, StateTracker& state_tracker_,
+                              Scheduler& scheduler_);
     ~RasterizerVulkan() override;
 
     void Draw(bool is_indexed, u32 instance_count) override;
@@ -90,18 +95,18 @@ public:
     void BindGraphicsUniformBuffer(size_t stage, u32 index, GPUVAddr gpu_addr, u32 size) override;
     void DisableGraphicsUniformBuffer(size_t stage, u32 index) override;
     void FlushAll() override;
-    void FlushRegion(VAddr addr, u64 size,
+    void FlushRegion(DAddr addr, u64 size,
                      VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    bool MustFlushRegion(VAddr addr, u64 size,
+    bool MustFlushRegion(DAddr addr, u64 size,
                          VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    VideoCore::RasterizerDownloadArea GetFlushArea(VAddr addr, u64 size) override;
+    VideoCore::RasterizerDownloadArea GetFlushArea(DAddr addr, u64 size) override;
-    void InvalidateRegion(VAddr addr, u64 size,
+    void InvalidateRegion(DAddr addr, u64 size,
                           VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
-    void InnerInvalidation(std::span<const std::pair<VAddr, std::size_t>> sequences) override;
+    void InnerInvalidation(std::span<const std::pair<DAddr, std::size_t>> sequences) override;
-    void OnCacheInvalidation(VAddr addr, u64 size) override;
+    void OnCacheInvalidation(DAddr addr, u64 size) override;
-    bool OnCPUWrite(VAddr addr, u64 size) override;
+    bool OnCPUWrite(DAddr addr, u64 size) override;
     void InvalidateGPUCache() override;
-    void UnmapMemory(VAddr addr, u64 size) override;
+    void UnmapMemory(DAddr addr, u64 size) override;
     void ModifyGPUMemory(size_t as_id, GPUVAddr addr, u64 size) override;
     void SignalFence(std::function<void()>&& func) override;
     void SyncOperation(std::function<void()>&& func) override;
@@ -109,7 +114,7 @@ public:
     void SignalReference() override;
     void ReleaseFences(bool force = true) override;
     void FlushAndInvalidateRegion(
-        VAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
+        DAddr addr, u64 size, VideoCommon::CacheType which = VideoCommon::CacheType::All) override;
     void WaitForIdle() override;
     void FragmentBarrier() override;
     void TiledCacheBarrier() override;
@@ -122,7 +127,7 @@ public:
     Tegra::Engines::AccelerateDMAInterface& AccessAccelerateDMA() override;
     void AccelerateInlineToMemory(GPUVAddr address, size_t copy_size,
                                   std::span<const u8> memory) override;
-    bool AccelerateDisplay(const Tegra::FramebufferConfig& config, VAddr framebuffer_addr,
+    bool AccelerateDisplay(const Tegra::FramebufferConfig& config, DAddr framebuffer_addr,
                            u32 pixel_stride) override;
     void LoadDiskResources(u64 title_id, std::stop_token stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;
@@ -176,6 +181,7 @@ private:
     void UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs);
 
     Tegra::GPU& gpu;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
 
     ScreenInfo& screen_info;
     const Device& device;

src/video_core/shader_cache.cpp

@@ -12,6 +12,7 @@
 #include "video_core/dirty_flags.h"
 #include "video_core/engines/kepler_compute.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/memory_manager.h"
 #include "video_core/shader_cache.h"
 #include "video_core/shader_environment.h"
@@ -34,7 +35,7 @@ void ShaderCache::SyncGuestHost() {
     RemovePendingShaders();
 }
 
-ShaderCache::ShaderCache(VideoCore::RasterizerInterface& rasterizer_) : rasterizer{rasterizer_} {}
+ShaderCache::ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory_) : device_memory{device_memory_} {}
 
 bool ShaderCache::RefreshStages(std::array<u64, 6>& unique_hashes) {
     auto& dirty{maxwell3d->dirty.flags};
@@ -132,7 +133,7 @@ void ShaderCache::Register(std::unique_ptr<ShaderInfo> data, VAddr addr, size_t
 
     storage.push_back(std::move(data));
 
-    rasterizer.UpdatePagesCachedCount(addr, size, 1);
+    device_memory.UpdatePagesCachedCount(addr, size, 1);
 }
 
 void ShaderCache::InvalidatePagesInRegion(VAddr addr, size_t size) {
@@ -209,7 +210,7 @@ void ShaderCache::UnmarkMemory(Entry* entry) {
 
     const VAddr addr = entry->addr_start;
     const size_t size = entry->addr_end - addr;
-    rasterizer.UpdatePagesCachedCount(addr, size, -1);
+    device_memory.UpdatePagesCachedCount(addr, size, -1);
 }
 
 void ShaderCache::RemoveShadersFromStorage(std::span<ShaderInfo*> removed_shaders) {

src/video_core/shader_cache.h

@@ -16,6 +16,7 @@
 #include "video_core/control/channel_state_cache.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/shader_environment.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 
 namespace Tegra {
 class MemoryManager;
@@ -77,7 +78,7 @@ protected:
         }
     };
 
-    explicit ShaderCache(VideoCore::RasterizerInterface& rasterizer_);
+    explicit ShaderCache(Tegra::MaxwellDeviceMemoryManager& device_memory);
 
     /// @brief Update the hashes and information of shader stages
     /// @param unique_hashes Shader hashes to store into when a stage is enabled
@@ -145,7 +146,7 @@ private:
     /// @brief Create a new shader entry and register it
     const ShaderInfo* MakeShaderInfo(GenericEnvironment& env, VAddr cpu_addr);
 
-    VideoCore::RasterizerInterface& rasterizer;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
 
     mutable std::mutex lookup_mutex;
     std::mutex invalidation_mutex;

src/video_core/texture_cache/texture_cache.h

@@ -8,10 +8,11 @@
 
 #include "common/alignment.h"
 #include "common/settings.h"
-#include "core/memory.h"
 #include "video_core/control/channel_state.h"
 #include "video_core/dirty_flags.h"
 #include "video_core/engines/kepler_compute.h"
+#include "video_core/guest_memory.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
 #include "video_core/texture_cache/image_view_base.h"
 #include "video_core/texture_cache/samples_helper.h"
 #include "video_core/texture_cache/texture_cache_base.h"
@@ -27,8 +28,8 @@ using VideoCore::Surface::SurfaceType;
 using namespace Common::Literals;
 
 template <class P>
-TextureCache<P>::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface& rasterizer_)
+TextureCache<P>::TextureCache(Runtime& runtime_, Tegra::MaxwellDeviceMemoryManager& device_memory_)
-    : runtime{runtime_}, rasterizer{rasterizer_} {
+    : runtime{runtime_}, device_memory{device_memory_} {
     // Configure null sampler
     TSCEntry sampler_descriptor{};
     sampler_descriptor.min_filter.Assign(Tegra::Texture::TextureFilter::Linear);
@@ -49,19 +50,19 @@ TextureCache<P>::TextureCache(Runtime& runtime_, VideoCore::RasterizerInterface&
     void(slot_samplers.insert(runtime, sampler_descriptor));
 
     if constexpr (HAS_DEVICE_MEMORY_INFO) {
-        const s64 device_memory = static_cast<s64>(runtime.GetDeviceLocalMemory());
+        const s64 device_local_memory = static_cast<s64>(runtime.GetDeviceLocalMemory());
-        const s64 min_spacing_expected = device_memory - 1_GiB;
+        const s64 min_spacing_expected = device_local_memory - 1_GiB;
-        const s64 min_spacing_critical = device_memory - 512_MiB;
+        const s64 min_spacing_critical = device_local_memory - 512_MiB;
-        const s64 mem_threshold = std::min(device_memory, TARGET_THRESHOLD);
+        const s64 mem_threshold = std::min(device_local_memory, TARGET_THRESHOLD);
         const s64 min_vacancy_expected = (6 * mem_threshold) / 10;
         const s64 min_vacancy_critical = (3 * mem_threshold) / 10;
         expected_memory = static_cast<u64>(
-            std::max(std::min(device_memory - min_vacancy_expected, min_spacing_expected),
+            std::max(std::min(device_local_memory - min_vacancy_expected, min_spacing_expected),
                      DEFAULT_EXPECTED_MEMORY));
         critical_memory = static_cast<u64>(
-            std::max(std::min(device_memory - min_vacancy_critical, min_spacing_critical),
+            std::max(std::min(device_local_memory - min_vacancy_critical, min_spacing_critical),
                      DEFAULT_CRITICAL_MEMORY));
-        minimum_memory = static_cast<u64>((device_memory - mem_threshold) / 2);
+        minimum_memory = static_cast<u64>((device_local_memory - mem_threshold) / 2);
     } else {
         expected_memory = DEFAULT_EXPECTED_MEMORY + 512_MiB;
         critical_memory = DEFAULT_CRITICAL_MEMORY + 1_GiB;
@@ -513,7 +514,7 @@ FramebufferId TextureCache<P>::GetFramebufferId(const RenderTargets& key) {
 }
 
 template <class P>
-void TextureCache<P>::WriteMemory(VAddr cpu_addr, size_t size) {
+void TextureCache<P>::WriteMemory(DAddr cpu_addr, size_t size) {
     ForEachImageInRegion(cpu_addr, size, [this](ImageId image_id, Image& image) {
         if (True(image.flags & ImageFlagBits::CpuModified)) {
             return;
@@ -526,7 +527,7 @@ void TextureCache<P>::WriteMemory(VAddr cpu_addr, size_t size) {
 }
 
 template <class P>
-void TextureCache<P>::DownloadMemory(VAddr cpu_addr, size_t size) {
+void TextureCache<P>::DownloadMemory(DAddr cpu_addr, size_t size) {
     boost::container::small_vector<ImageId, 16> images;
     ForEachImageInRegion(cpu_addr, size, [&images](ImageId image_id, ImageBase& image) {
         if (!image.IsSafeDownload()) {
@@ -553,7 +554,7 @@ void TextureCache<P>::DownloadMemory(VAddr cpu_addr, size_t size) {
 }
 
 template <class P>
-std::optional<VideoCore::RasterizerDownloadArea> TextureCache<P>::GetFlushArea(VAddr cpu_addr,
+std::optional<VideoCore::RasterizerDownloadArea> TextureCache<P>::GetFlushArea(DAddr cpu_addr,
                                                                                u64 size) {
     std::optional<VideoCore::RasterizerDownloadArea> area{};
     ForEachImageInRegion(cpu_addr, size, [&](ImageId, ImageBase& image) {
@@ -579,7 +580,7 @@ std::optional<VideoCore::RasterizerDownloadArea> TextureCache<P>::GetFlushArea(V
 }
 
 template <class P>
-void TextureCache<P>::UnmapMemory(VAddr cpu_addr, size_t size) {
+void TextureCache<P>::UnmapMemory(DAddr cpu_addr, size_t size) {
     boost::container::small_vector<ImageId, 16> deleted_images;
     ForEachImageInRegion(cpu_addr, size, [&](ImageId id, Image&) { deleted_images.push_back(id); });
     for (const ImageId id : deleted_images) {
@@ -713,7 +714,7 @@ bool TextureCache<P>::BlitImage(const Tegra::Engines::Fermi2D::Surface& dst,
 
 template <class P>
 typename P::ImageView* TextureCache<P>::TryFindFramebufferImageView(
-    const Tegra::FramebufferConfig& config, VAddr cpu_addr) {
+    const Tegra::FramebufferConfig& config, DAddr cpu_addr) {
     // TODO: Properly implement this
     const auto it = page_table.find(cpu_addr >> YUZU_PAGEBITS);
     if (it == page_table.end()) {
@@ -940,7 +941,7 @@ bool TextureCache<P>::IsRescaling(const ImageViewBase& image_view) const noexcep
 }
 
 template <class P>
-bool TextureCache<P>::IsRegionGpuModified(VAddr addr, size_t size) {
+bool TextureCache<P>::IsRegionGpuModified(DAddr addr, size_t size) {
     bool is_modified = false;
     ForEachImageInRegion(addr, size, [&is_modified](ImageId, ImageBase& image) {
         if (False(image.flags & ImageFlagBits::GpuModified)) {
@@ -1059,7 +1060,7 @@ void TextureCache<P>::UploadImageContents(Image& image, StagingBuffer& staging)
         return;
     }
 
-    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> swizzle_data(
+    Tegra::Memory::GpuGuestMemory<u8, Tegra::Memory::GuestMemoryFlags::UnsafeRead> swizzle_data(
         *gpu_memory, gpu_addr, image.guest_size_bytes, &swizzle_data_buffer);
 
     if (True(image.flags & ImageFlagBits::Converted)) {
@@ -1124,7 +1125,7 @@ ImageId TextureCache<P>::FindOrInsertImage(const ImageInfo& info, GPUVAddr gpu_a
 template <class P>
 ImageId TextureCache<P>::FindImage(const ImageInfo& info, GPUVAddr gpu_addr,
                                    RelaxedOptions options) {
-    std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+    std::optional<DAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
     if (!cpu_addr) {
         cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr, CalculateGuestSizeInBytes(info));
         if (!cpu_addr) {
@@ -1265,7 +1266,7 @@ void TextureCache<P>::QueueAsyncDecode(Image& image, ImageId image_id) {
 
     static Common::ScratchBuffer<u8> local_unswizzle_data_buffer;
     local_unswizzle_data_buffer.resize_destructive(image.unswizzled_size_bytes);
-    Core::Memory::GpuGuestMemory<u8, Core::Memory::GuestMemoryFlags::UnsafeRead> swizzle_data(
+    Tegra::Memory::GpuGuestMemory<u8, Tegra::Memory::GuestMemoryFlags::UnsafeRead> swizzle_data(
         *gpu_memory, image.gpu_addr, image.guest_size_bytes, &swizzle_data_buffer);
 
     auto copies = UnswizzleImage(*gpu_memory, image.gpu_addr, image.info, swizzle_data,
@@ -1339,14 +1340,14 @@ bool TextureCache<P>::ScaleDown(Image& image) {
 template <class P>
 ImageId TextureCache<P>::InsertImage(const ImageInfo& info, GPUVAddr gpu_addr,
                                      RelaxedOptions options) {
-    std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+    std::optional<DAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
     if (!cpu_addr) {
         const auto size = CalculateGuestSizeInBytes(info);
         cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr, size);
         if (!cpu_addr) {
-            const VAddr fake_addr = ~(1ULL << 40ULL) + virtual_invalid_space;
+            const DAddr fake_addr = ~(1ULL << 40ULL) + virtual_invalid_space;
             virtual_invalid_space += Common::AlignUp(size, 32);
-            cpu_addr = std::optional<VAddr>(fake_addr);
+            cpu_addr = std::optional<DAddr>(fake_addr);
         }
     }
     ASSERT_MSG(cpu_addr, "Tried to insert an image to an invalid gpu_addr=0x{:x}", gpu_addr);
@@ -1362,7 +1363,7 @@ ImageId TextureCache<P>::InsertImage(const ImageInfo& info, GPUVAddr gpu_addr,
 }
 
 template <class P>
-ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VAddr cpu_addr) {
+ImageId TextureCache<P>::JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, DAddr cpu_addr) {
     ImageInfo new_info = info;
     const size_t size_bytes = CalculateGuestSizeInBytes(new_info);
     const bool broken_views = runtime.HasBrokenTextureViewFormats();
@@ -1650,7 +1651,7 @@ std::optional<typename TextureCache<P>::BlitImages> TextureCache<P>::GetBlitImag
 
 template <class P>
 ImageId TextureCache<P>::FindDMAImage(const ImageInfo& info, GPUVAddr gpu_addr) {
-    std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+    std::optional<DAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
     if (!cpu_addr) {
         cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr, CalculateGuestSizeInBytes(info));
         if (!cpu_addr) {
@@ -1780,7 +1781,7 @@ ImageViewId TextureCache<P>::FindRenderTargetView(const ImageInfo& info, GPUVAdd
 
 template <class P>
 template <typename Func>
-void TextureCache<P>::ForEachImageInRegion(VAddr cpu_addr, size_t size, Func&& func) {
+void TextureCache<P>::ForEachImageInRegion(DAddr cpu_addr, size_t size, Func&& func) {
     using FuncReturn = typename std::invoke_result<Func, ImageId, Image&>::type;
     static constexpr bool BOOL_BREAK = std::is_same_v<FuncReturn, bool>;
     boost::container::small_vector<ImageId, 32> images;
@@ -1924,11 +1925,11 @@ void TextureCache<P>::ForEachSparseImageInRegion(GPUVAddr gpu_addr, size_t size,
 template <class P>
 template <typename Func>
 void TextureCache<P>::ForEachSparseSegment(ImageBase& image, Func&& func) {
-    using FuncReturn = typename std::invoke_result<Func, GPUVAddr, VAddr, size_t>::type;
+    using FuncReturn = typename std::invoke_result<Func, GPUVAddr, DAddr, size_t>::type;
     static constexpr bool RETURNS_BOOL = std::is_same_v<FuncReturn, bool>;
     const auto segments = gpu_memory->GetSubmappedRange(image.gpu_addr, image.guest_size_bytes);
     for (const auto& [gpu_addr, size] : segments) {
-        std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+        std::optional<DAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
         ASSERT(cpu_addr);
         if constexpr (RETURNS_BOOL) {
             if (func(gpu_addr, *cpu_addr, size)) {
@@ -1980,7 +1981,7 @@ void TextureCache<P>::RegisterImage(ImageId image_id) {
     }
     boost::container::small_vector<ImageViewId, 16> sparse_maps;
     ForEachSparseSegment(
-        image, [this, image_id, &sparse_maps](GPUVAddr gpu_addr, VAddr cpu_addr, size_t size) {
+        image, [this, image_id, &sparse_maps](GPUVAddr gpu_addr, DAddr cpu_addr, size_t size) {
             auto map_id = slot_map_views.insert(gpu_addr, cpu_addr, size, image_id);
             ForEachCPUPage(cpu_addr, size,
                            [this, map_id](u64 page) { page_table[page].push_back(map_id); });
@@ -2048,7 +2049,7 @@ void TextureCache<P>::UnregisterImage(ImageId image_id) {
     auto& sparse_maps = it->second;
     for (auto& map_view_id : sparse_maps) {
         const auto& map_range = slot_map_views[map_view_id];
-        const VAddr cpu_addr = map_range.cpu_addr;
+        const DAddr cpu_addr = map_range.cpu_addr;
         const std::size_t size = map_range.size;
         ForEachCPUPage(cpu_addr, size, [this, image_id](u64 page) {
             const auto page_it = page_table.find(page);
@@ -2080,7 +2081,7 @@ void TextureCache<P>::TrackImage(ImageBase& image, ImageId image_id) {
     ASSERT(False(image.flags & ImageFlagBits::Tracked));
     image.flags |= ImageFlagBits::Tracked;
     if (False(image.flags & ImageFlagBits::Sparse)) {
-        rasterizer.UpdatePagesCachedCount(image.cpu_addr, image.guest_size_bytes, 1);
+        device_memory.UpdatePagesCachedCount(image.cpu_addr, image.guest_size_bytes, 1);
         return;
     }
     if (True(image.flags & ImageFlagBits::Registered)) {
@@ -2089,15 +2090,15 @@ void TextureCache<P>::TrackImage(ImageBase& image, ImageId image_id) {
         auto& sparse_maps = it->second;
         for (auto& map_view_id : sparse_maps) {
             const auto& map = slot_map_views[map_view_id];
-            const VAddr cpu_addr = map.cpu_addr;
+            const DAddr cpu_addr = map.cpu_addr;
             const std::size_t size = map.size;
-            rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
+            device_memory.UpdatePagesCachedCount(cpu_addr, size, 1);
         }
         return;
     }
     ForEachSparseSegment(image,
-                         [this]([[maybe_unused]] GPUVAddr gpu_addr, VAddr cpu_addr, size_t size) {
+                         [this]([[maybe_unused]] GPUVAddr gpu_addr, DAddr cpu_addr, size_t size) {
-                             rasterizer.UpdatePagesCachedCount(cpu_addr, size, 1);
+                             device_memory.UpdatePagesCachedCount(cpu_addr, size, 1);
                          });
 }
 
@@ -2106,7 +2107,7 @@ void TextureCache<P>::UntrackImage(ImageBase& image, ImageId image_id) {
     ASSERT(True(image.flags & ImageFlagBits::Tracked));
     image.flags &= ~ImageFlagBits::Tracked;
     if (False(image.flags & ImageFlagBits::Sparse)) {
-        rasterizer.UpdatePagesCachedCount(image.cpu_addr, image.guest_size_bytes, -1);
+        device_memory.UpdatePagesCachedCount(image.cpu_addr, image.guest_size_bytes, -1);
         return;
     }
     ASSERT(True(image.flags & ImageFlagBits::Registered));
@@ -2115,9 +2116,9 @@ void TextureCache<P>::UntrackImage(ImageBase& image, ImageId image_id) {
     auto& sparse_maps = it->second;
     for (auto& map_view_id : sparse_maps) {
         const auto& map = slot_map_views[map_view_id];
-        const VAddr cpu_addr = map.cpu_addr;
+        const DAddr cpu_addr = map.cpu_addr;
         const std::size_t size = map.size;
-        rasterizer.UpdatePagesCachedCount(cpu_addr, size, -1);
+        device_memory.UpdatePagesCachedCount(cpu_addr, size, -1);
     }
 }
 

src/video_core/texture_cache/texture_cache_base.h

@@ -36,9 +36,11 @@
 #include "video_core/texture_cache/types.h"
 #include "video_core/textures/texture.h"
 
-namespace Tegra::Control {
+namespace Tegra {
+namespace Control {
 struct ChannelState;
 }
+} // namespace Tegra
 
 namespace VideoCommon {
 
@@ -126,7 +128,7 @@ class TextureCache : public VideoCommon::ChannelSetupCaches<TextureCacheChannelI
     };
 
 public:
-    explicit TextureCache(Runtime&, VideoCore::RasterizerInterface&);
+    explicit TextureCache(Runtime&, Tegra::MaxwellDeviceMemoryManager&);
 
     /// Notify the cache that a new frame has been queued
     void TickFrame();
@@ -190,15 +192,15 @@ public:
     Framebuffer* GetFramebuffer();
 
     /// Mark images in a range as modified from the CPU
-    void WriteMemory(VAddr cpu_addr, size_t size);
+    void WriteMemory(DAddr cpu_addr, size_t size);
 
     /// Download contents of host images to guest memory in a region
-    void DownloadMemory(VAddr cpu_addr, size_t size);
+    void DownloadMemory(DAddr cpu_addr, size_t size);
 
-    std::optional<VideoCore::RasterizerDownloadArea> GetFlushArea(VAddr cpu_addr, u64 size);
+    std::optional<VideoCore::RasterizerDownloadArea> GetFlushArea(DAddr cpu_addr, u64 size);
 
     /// Remove images in a region
-    void UnmapMemory(VAddr cpu_addr, size_t size);
+    void UnmapMemory(DAddr cpu_addr, size_t size);
 
     /// Remove images in a region
     void UnmapGPUMemory(size_t as_id, GPUVAddr gpu_addr, size_t size);
@@ -210,7 +212,7 @@ public:
 
     /// Try to find a cached image view in the given CPU address
     [[nodiscard]] ImageView* TryFindFramebufferImageView(const Tegra::FramebufferConfig& config,
-                                                         VAddr cpu_addr);
+                                                         DAddr cpu_addr);
 
     /// Return true when there are uncommitted images to be downloaded
     [[nodiscard]] bool HasUncommittedFlushes() const noexcept;
@@ -235,7 +237,7 @@ public:
                                  GPUVAddr address = 0, size_t size = 0);
 
     /// Return true when a CPU region is modified from the GPU
-    [[nodiscard]] bool IsRegionGpuModified(VAddr addr, size_t size);
+    [[nodiscard]] bool IsRegionGpuModified(DAddr addr, size_t size);
 
     [[nodiscard]] bool IsRescaling() const noexcept;
 
@@ -252,7 +254,7 @@ public:
 private:
     /// Iterate over all page indices in a range
     template <typename Func>
-    static void ForEachCPUPage(VAddr addr, size_t size, Func&& func) {
+    static void ForEachCPUPage(DAddr addr, size_t size, Func&& func) {
         static constexpr bool RETURNS_BOOL = std::is_same_v<std::invoke_result<Func, u64>, bool>;
         const u64 page_end = (addr + size - 1) >> YUZU_PAGEBITS;
         for (u64 page = addr >> YUZU_PAGEBITS; page <= page_end; ++page) {
@@ -326,7 +328,7 @@ private:
 
     /// Create a new image and join perfectly matching existing images
     /// Remove joined images from the cache
-    [[nodiscard]] ImageId JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, VAddr cpu_addr);
+    [[nodiscard]] ImageId JoinImages(const ImageInfo& info, GPUVAddr gpu_addr, DAddr cpu_addr);
 
     [[nodiscard]] ImageId FindDMAImage(const ImageInfo& info, GPUVAddr gpu_addr);
 
@@ -349,7 +351,7 @@ private:
 
     /// Iterates over all the images in a region calling func
     template <typename Func>
-    void ForEachImageInRegion(VAddr cpu_addr, size_t size, Func&& func);
+    void ForEachImageInRegion(DAddr cpu_addr, size_t size, Func&& func);
 
     template <typename Func>
     void ForEachImageInRegionGPU(size_t as_id, GPUVAddr gpu_addr, size_t size, Func&& func);
@@ -421,7 +423,7 @@ private:
 
     Runtime& runtime;
 
-    VideoCore::RasterizerInterface& rasterizer;
+    Tegra::MaxwellDeviceMemoryManager& device_memory;
     std::deque<TextureCacheGPUMap> gpu_page_table_storage;
 
     RenderTargets render_targets;
@@ -432,7 +434,7 @@ private:
     std::unordered_map<u64, std::vector<ImageId>, Common::IdentityHash<u64>> sparse_page_table;
     std::unordered_map<ImageId, boost::container::small_vector<ImageViewId, 16>> sparse_views;
 
-    VAddr virtual_invalid_space{};
+    DAddr virtual_invalid_space{};
 
     bool has_deleted_images = false;
     bool is_rescaling = false;

src/video_core/texture_cache/util.cpp

@@ -23,6 +23,7 @@
 #include "core/memory.h"
 #include "video_core/compatible_formats.h"
 #include "video_core/engines/maxwell_3d.h"
+#include "video_core/guest_memory.h"
 #include "video_core/memory_manager.h"
 #include "video_core/surface.h"
 #include "video_core/texture_cache/decode_bc.h"
@@ -552,7 +553,8 @@ void SwizzleBlockLinearImage(Tegra::MemoryManager& gpu_memory, GPUVAddr gpu_addr
     for (s32 layer = 0; layer < info.resources.layers; ++layer) {
         const std::span<const u8> src = input.subspan(host_offset);
         {
-            Core::Memory::GpuGuestMemoryScoped<u8, Core::Memory::GuestMemoryFlags::UnsafeReadWrite>
+            Tegra::Memory::GpuGuestMemoryScoped<u8,
+                                                Tegra::Memory::GuestMemoryFlags::UnsafeReadWrite>
                 dst(gpu_memory, gpu_addr + guest_offset, subresource_size, &tmp_buffer);
 
             SwizzleTexture(dst, src, bytes_per_block, num_tiles.width, num_tiles.height,

src/video_core/video_core.cpp

@@ -6,6 +6,8 @@
 #include "common/logging/log.h"
 #include "common/settings.h"
 #include "core/core.h"
+#include "video_core/host1x/gpu_device_memory_manager.h"
+#include "video_core/host1x/host1x.h"
 #include "video_core/renderer_base.h"
 #include "video_core/renderer_null/renderer_null.h"
 #include "video_core/renderer_opengl/renderer_opengl.h"
@@ -18,18 +20,17 @@ std::unique_ptr<VideoCore::RendererBase> CreateRenderer(
     Core::System& system, Core::Frontend::EmuWindow& emu_window, Tegra::GPU& gpu,
     std::unique_ptr<Core::Frontend::GraphicsContext> context) {
     auto& telemetry_session = system.TelemetrySession();
-    auto& cpu_memory = system.ApplicationMemory();
+    auto& device_memory = system.Host1x().MemoryManager();
 
     switch (Settings::values.renderer_backend.GetValue()) {
     case Settings::RendererBackend::OpenGL:
-        return std::make_unique<OpenGL::RendererOpenGL>(telemetry_session, emu_window, cpu_memory,
+        return std::make_unique<OpenGL::RendererOpenGL>(telemetry_session, emu_window,
-                                                        gpu, std::move(context));
+                                                        device_memory, gpu, std::move(context));
     case Settings::RendererBackend::Vulkan:
-        return std::make_unique<Vulkan::RendererVulkan>(telemetry_session, emu_window, cpu_memory,
+        return std::make_unique<Vulkan::RendererVulkan>(telemetry_session, emu_window,
-                                                        gpu, std::move(context));
+                                                        device_memory, gpu, std::move(context));
     case Settings::RendererBackend::Null:
-        return std::make_unique<Null::RendererNull>(emu_window, cpu_memory, gpu,
+        return std::make_unique<Null::RendererNull>(emu_window, gpu, std::move(context));
-                                                    std::move(context));
     default:
         return nullptr;
     }