9 files changed, 42 insertions, 50 deletions

diff --git a/src/core/device_memory_manager.inc b/src/core/device_memory_manager.inc
index 138eb5017..4f883cece 100644
--- a/src/core/device_memory_manager.inc
+++ b/src/core/device_memory_manager.inc
@@ -217,9 +217,6 @@ DeviceMemoryManager<Traits>::DeviceMemoryManager(const DeviceMemory& device_memo
       cpu_backing_address(device_as_size >> Memory::YUZU_PAGEBITS) {
     impl = std::make_unique<DeviceMemoryManagerAllocator<Traits>>();
     cached_pages = std::make_unique<CachedPages>();
-    for (size_t i = 0; i < 1ULL << (33 - 12); i++) {
-        compressed_device_addr[i] = 0;
-    }
 }
 
 template <typename Traits>
@@ -517,7 +514,7 @@ void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size
     u64 cache_begin = 0;
     u64 uncache_bytes = 0;
     u64 cache_bytes = 0;
-    const auto* MarkRegionCaching = &DeviceMemoryManager<Traits>::DeviceMethods::MarkRegionCaching;
+    const auto MarkRegionCaching = &DeviceMemoryManager<Traits>::DeviceMethods::MarkRegionCaching;
 
     std::atomic_thread_fence(std::memory_order_acquire);
     const size_t page_end = Common::DivCeil(addr + size, Memory::YUZU_PAGESIZE);
@@ -577,4 +574,4 @@ void DeviceMemoryManager<Traits>::UpdatePagesCachedCount(DAddr addr, size_t size
     }
 }
 
-} // namespace Core
\ No newline at end of file
+} // namespace Core
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
index 0b6aa9993..a50577c75 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec_common.cpp
@@ -8,6 +8,7 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
+#include "core/hle/kernel/k_process.h"
 #include "core/hle/service/nvdrv/core/container.h"
 #include "core/hle/service/nvdrv/core/nvmap.h"
 #include "core/hle/service/nvdrv/core/syncpoint_manager.h"
@@ -109,7 +110,7 @@ NvResult nvhost_nvdec_common::Submit(IoctlSubmit& params, std::span<u8> data, De
         ASSERT_OR_EXECUTE(object, return NvResult::InvalidState;);
         Tegra::ChCommandHeaderList cmdlist(cmd_buffer.word_count);
         session->process->GetMemory().ReadBlock(object->address + cmd_buffer.offset, cmdlist.data(),
-                                             cmdlist.size() * sizeof(u32));
+                                                cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(core.Host1xDeviceFile().fd_to_id[fd], cmdlist);
     }
     // Some games expect command_buffers to be written back
@@ -135,7 +136,8 @@ NvResult nvhost_nvdec_common::GetWaitbase(IoctlGetWaitbase& params) {
     return NvResult::Success;
 }
 
-NvResult nvhost_nvdec_common::MapBuffer(IoctlMapBuffer& params, std::span<MapBufferEntry> entries, DeviceFD fd) {
+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++) {
         DAddr pin_address = nvmap.PinHandle(entries[i].map_handle, sessions[fd], true);
diff --git a/src/core/memory.cpp b/src/core/memory.cpp
index f126840cb..1c218566f 100644
--- a/src/core/memory.cpp
+++ b/src/core/memory.cpp
@@ -44,8 +44,7 @@ bool AddressSpaceContains(const Common::PageTable& table, const Common::ProcessA
 // from outside classes. This also allows modification to the internals of the memory
 // subsystem without needing to rebuild all files that make use of the memory interface.
 struct Memory::Impl {
-    explicit Impl(Core::System& system_)
-        : system{system_} {}
+    explicit Impl(Core::System& system_) : system{system_} {}
 
     void SetCurrentPageTable(Kernel::KProcess& process) {
         current_page_table = &process.GetPageTable().GetImpl();
@@ -640,18 +639,6 @@ struct Memory::Impl {
         LOG_DEBUG(HW_Memory, "Mapping {:016X} onto {:016X}-{:016X}", GetInteger(target),
                   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()) {
-            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}",
                    base + page_table.pointers.size());
@@ -823,8 +810,7 @@ struct Memory::Impl {
         }
         const size_t core = system.GetCurrentHostThreadID();
         auto& current_area = rasterizer_read_areas[core];
-        gpu_device_memory->ApplyOpOnPointer(
-            p, scratch_buffers[core], [&](DAddr address) {
+        gpu_device_memory->ApplyOpOnPointer(p, scratch_buffers[core], [&](DAddr address) {
             const DAddr end_address = address + size;
             if (current_area.start_address <= address && end_address <= current_area.end_address)
                 [[likely]] {
@@ -852,8 +838,7 @@ struct Memory::Impl {
                 sys_core_guard.unlock();
             }
         });
-        gpu_device_memory->ApplyOpOnPointer(
-            p, scratch_buffers[core], [&](DAddr address) {
+        gpu_device_memory->ApplyOpOnPointer(p, scratch_buffers[core], [&](DAddr address) {
             auto& current_area = rasterizer_write_areas[core];
             PAddr subaddress = address >> YUZU_PAGEBITS;
             bool do_collection = current_area.last_address == subaddress;
@@ -872,12 +857,25 @@ struct Memory::Impl {
         PAddr last_address;
     };
 
-    void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) {
-        system.GPU().InvalidateRegion(GetInteger(dest_addr), size);
-    }
-
-    void FlushRegion(Common::ProcessAddress dest_addr, size_t size) {
-        system.GPU().FlushRegion(GetInteger(dest_addr), size);
+    void InvalidateGPUMemory(u8* p, size_t size) {
+        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.
+        if (!gpu_device_memory) [[unlikely]] {
+            gpu_device_memory = &system.Host1x().MemoryManager();
+        }
+        // Guard on sys_core;
+        if (core == sys_core) [[unlikely]] {
+            sys_core_guard.lock();
+        }
+        SCOPE_EXIT({
+            if (core == sys_core) [[unlikely]] {
+                sys_core_guard.unlock();
+            }
+        });
+        auto& gpu = system.GPU();
+        gpu_device_memory->ApplyOpOnPointer(
+            p, scratch_buffers[core], [&](DAddr address) { gpu.InvalidateRegion(address, size); });
     }
 
     Core::System& system;
@@ -1081,14 +1079,6 @@ void Memory::MarkRegionDebug(Common::ProcessAddress vaddr, u64 size, bool debug)
     impl->MarkRegionDebug(GetInteger(vaddr), size, debug);
 }
 
-void Memory::InvalidateRegion(Common::ProcessAddress dest_addr, size_t size) {
-    impl->InvalidateRegion(dest_addr, size);
-}
-
-void Memory::FlushRegion(Common::ProcessAddress dest_addr, size_t size) {
-    impl->FlushRegion(dest_addr, size);
-}
-
 bool Memory::InvalidateNCE(Common::ProcessAddress vaddr, size_t size) {
     [[maybe_unused]] bool mapped = true;
     [[maybe_unused]] bool rasterizer = false;
@@ -1100,10 +1090,10 @@ bool Memory::InvalidateNCE(Common::ProcessAddress vaddr, size_t size) {
                       GetInteger(vaddr));
             mapped = false;
         },
-        [&] {
-            impl->system.GPU().InvalidateRegion(GetInteger(vaddr), size);
-            rasterizer = true;
-        });
+        [&] { rasterizer = true; });
+    if (rasterizer) {
+        impl->InvalidateGPUMemory(ptr, size);
+    }
 
 #ifdef __linux__
     if (!rasterizer && mapped) {
diff --git a/src/core/memory.h b/src/core/memory.h
index 47ca6a35a..9d29cfd3f 100644
--- a/src/core/memory.h
+++ b/src/core/memory.h
@@ -486,10 +486,10 @@ public:
     void MarkRegionDebug(Common::ProcessAddress vaddr, u64 size, bool debug);
 
     void SetGPUDirtyManagers(std::span<Core::GPUDirtyMemoryManager> managers);
-    void InvalidateRegion(Common::ProcessAddress dest_addr, size_t size);
+
     bool InvalidateNCE(Common::ProcessAddress vaddr, size_t size);
+    
     bool InvalidateSeparateHeap(void* fault_address);
-    void FlushRegion(Common::ProcessAddress dest_addr, size_t size);
 
 private:
     Core::System& system;
diff --git a/src/video_core/host1x/gpu_device_memory_manager.h b/src/video_core/host1x/gpu_device_memory_manager.h
index 6c7858848..9ccd84b9a 100644
--- a/src/video_core/host1x/gpu_device_memory_manager.h
+++ b/src/video_core/host1x/gpu_device_memory_manager.h
@@ -17,7 +17,7 @@ struct MaxwellDeviceTraits {
     static constexpr bool supports_pinning = false;
     static constexpr size_t device_virtual_bits = 34;
     using DeviceInterface = typename VideoCore::RasterizerInterface;
-    using DeviceMethods = typename MaxwellDeviceMethods;
+    using DeviceMethods = MaxwellDeviceMethods;
 };
 
 using MaxwellDeviceMemoryManager = Core::DeviceMemoryManager<MaxwellDeviceTraits>;
diff --git a/src/video_core/host1x/host1x.cpp b/src/video_core/host1x/host1x.cpp
index b7f9a08cf..c4c7a5883 100644
--- a/src/video_core/host1x/host1x.cpp
+++ b/src/video_core/host1x/host1x.cpp
@@ -13,6 +13,8 @@ Host1x::Host1x(Core::System& system_)
       memory_manager(system.DeviceMemory()), gmmu_manager{system, memory_manager, 32, 12},
       allocator{std::make_unique<Common::FlatAllocator<u32, 0, 32>>(1 << 12)} {}
 
+Host1x::~Host1x() = default;
+
 } // namespace Host1x
 
 } // namespace Tegra
diff --git a/src/video_core/host1x/host1x.h b/src/video_core/host1x/host1x.h
index 13c37e6b4..d72d97b7b 100644
--- a/src/video_core/host1x/host1x.h
+++ b/src/video_core/host1x/host1x.h
@@ -21,6 +21,7 @@ namespace Host1x {
 class Host1x {
 public:
     explicit Host1x(Core::System& system);
+    ~Host1x();
 
     SyncpointManager& GetSyncpointManager() {
         return syncpoint_manager;
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index 6b2cd7efb..00d64dcce 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -68,7 +68,7 @@ public:
         if (!address) {
             return {};
         }
-        return memory.GetPointer(*address);
+        return memory.GetPointer<T>(*address);
     }
 
     template <typename T>
diff --git a/src/video_core/query_cache/query_cache.h b/src/video_core/query_cache/query_cache.h
index 508afb10a..b5e90cf8c 100644
--- a/src/video_core/query_cache/query_cache.h
+++ b/src/video_core/query_cache/query_cache.h
@@ -256,8 +256,8 @@ 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->device_memory.GetPointer<u8>(cpu_addr);
-    u8* pointer_timestamp = impl->device_memory.GetPointer<u8>(cpu_addr + 8);
+    u8* pointer = impl->device_memory.template GetPointer<u8>(cpu_addr);
+    u8* pointer_timestamp = impl->device_memory.template 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] {
@@ -561,7 +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->device_memory.GetPointer<u8>(query_base->guest_address);
+        auto* ptr = impl->device_memory.template 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;