renderer_opengl: Fix rebase mistake

Removes the sRGB hack of tracking if a frame used an sRGB rendertarget
to apply at least once to blit the final texture as sRGB. Instead of
doing this apply sRGB if the presented image has sRGB.

Also enable sRGB by default on Maxwell3D registers as some games seem to
assume this.

src/core/core.cpp

@@ -104,7 +104,8 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
     return vfs->OpenFile(path, FileSys::Mode::Read);
 }
 struct System::Impl {
-    explicit Impl(System& system) : kernel{system}, cpu_core_manager{system}, reporter{system} {}
+    explicit Impl(System& system)
+        : kernel{system}, cpu_core_manager{system}, applet_manager{system}, reporter{system} {}
 
     Cpu& CurrentCpuCore() {
         return cpu_core_manager.GetCurrentCore();

src/core/hle/kernel/vm_manager.cpp

@@ -296,12 +296,6 @@ ResultVal<VAddr> VMManager::SetHeapSize(u64 size) {
 }
 
 ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
-    const auto end_addr = target + size;
-    const auto last_addr = end_addr - 1;
-    VAddr cur_addr = target;
-
-    ResultCode result = RESULT_SUCCESS;
-
     // Check how much memory we've already mapped.
     const auto mapped_size_result = SizeOfAllocatedVMAsInRange(target, size);
     if (mapped_size_result.Failed()) {
@@ -324,13 +318,16 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
 
     // Keep track of the memory regions we unmap.
     std::vector<std::pair<u64, u64>> mapped_regions;
+    ResultCode result = RESULT_SUCCESS;
 
     // Iterate, trying to map memory.
     {
-        cur_addr = target;
+        const auto end_addr = target + size;
+        const auto last_addr = end_addr - 1;
+        VAddr cur_addr = target;
 
         auto iter = FindVMA(target);
-        ASSERT_MSG(iter != vma_map.end(), "MapPhysicalMemory iter != end");
+        ASSERT(iter != vma_map.end());
 
         while (true) {
             const auto& vma = iter->second;
@@ -342,7 +339,7 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
             const auto map_size = std::min(end_addr - cur_addr, vma_end - cur_addr);
             if (vma.state == MemoryState::Unmapped) {
                 const auto map_res =
-                    MapMemoryBlock(cur_addr, std::make_shared<PhysicalMemory>(map_size, 0), 0,
+                    MapMemoryBlock(cur_addr, std::make_shared<PhysicalMemory>(map_size), 0,
                                    map_size, MemoryState::Heap, VMAPermission::ReadWrite);
                 result = map_res.Code();
                 if (result.IsError()) {
@@ -360,7 +357,7 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
             // Advance to the next block.
             cur_addr = vma_end;
             iter = FindVMA(cur_addr);
-            ASSERT_MSG(iter != vma_map.end(), "MapPhysicalMemory iter != end");
+            ASSERT(iter != vma_map.end());
         }
     }
 
@@ -368,7 +365,7 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
     if (result.IsError()) {
         for (const auto [unmap_address, unmap_size] : mapped_regions) {
             ASSERT_MSG(UnmapRange(unmap_address, unmap_size).IsSuccess(),
-                       "MapPhysicalMemory un-map on error");
+                       "Failed to unmap memory range.");
         }
 
         return result;
@@ -381,12 +378,6 @@ ResultCode VMManager::MapPhysicalMemory(VAddr target, u64 size) {
 }
 
 ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
-    const auto end_addr = target + size;
-    const auto last_addr = end_addr - 1;
-    VAddr cur_addr = target;
-
-    ResultCode result = RESULT_SUCCESS;
-
     // Check how much memory is currently mapped.
     const auto mapped_size_result = SizeOfUnmappablePhysicalMemoryInRange(target, size);
     if (mapped_size_result.Failed()) {
@@ -401,13 +392,16 @@ ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
 
     // Keep track of the memory regions we unmap.
     std::vector<std::pair<u64, u64>> unmapped_regions;
+    ResultCode result = RESULT_SUCCESS;
 
     // Try to unmap regions.
     {
-        cur_addr = target;
+        const auto end_addr = target + size;
+        const auto last_addr = end_addr - 1;
+        VAddr cur_addr = target;
 
         auto iter = FindVMA(target);
-        ASSERT_MSG(iter != vma_map.end(), "UnmapPhysicalMemory iter != end");
+        ASSERT(iter != vma_map.end());
 
         while (true) {
             const auto& vma = iter->second;
@@ -434,7 +428,7 @@ ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
             // Advance to the next block.
             cur_addr = vma_end;
             iter = FindVMA(cur_addr);
-            ASSERT_MSG(iter != vma_map.end(), "UnmapPhysicalMemory iter != end");
+            ASSERT(iter != vma_map.end());
         }
     }
 
@@ -443,10 +437,12 @@ ResultCode VMManager::UnmapPhysicalMemory(VAddr target, u64 size) {
     if (result.IsError()) {
         for (const auto [map_address, map_size] : unmapped_regions) {
             const auto remap_res =
-                MapMemoryBlock(map_address, std::make_shared<PhysicalMemory>(map_size, 0), 0,
-                               map_size, MemoryState::Heap, VMAPermission::None);
-            ASSERT_MSG(remap_res.Succeeded(), "UnmapPhysicalMemory re-map on error");
+                MapMemoryBlock(map_address, std::make_shared<PhysicalMemory>(map_size), 0, map_size,
+                               MemoryState::Heap, VMAPermission::None);
+            ASSERT_MSG(remap_res.Succeeded(), "Failed to remap a memory block.");
         }
+
+        return result;
     }
 
     // Update mapped amount
@@ -757,20 +753,26 @@ void VMManager::MergeAdjacentVMA(VirtualMemoryArea& left, const VirtualMemoryAre
     // Always merge allocated memory blocks, even when they don't share the same backing block.
     if (left.type == VMAType::AllocatedMemoryBlock &&
         (left.backing_block != right.backing_block || left.offset + left.size != right.offset)) {
+        const auto right_begin = right.backing_block->begin() + right.offset;
+        const auto right_end = right_begin + right.size;
+
         // Check if we can save work.
         if (left.offset == 0 && left.size == left.backing_block->size()) {
             // Fast case: left is an entire backing block.
-            left.backing_block->insert(left.backing_block->end(),
-                                       right.backing_block->begin() + right.offset,
-                                       right.backing_block->begin() + right.offset + right.size);
+            left.backing_block->insert(left.backing_block->end(), right_begin, right_end);
         } else {
             // Slow case: make a new memory block for left and right.
+            const auto left_begin = left.backing_block->begin() + left.offset;
+            const auto left_end = left_begin + left.size;
+            const auto left_size = static_cast<std::size_t>(std::distance(left_begin, left_end));
+            const auto right_size = static_cast<std::size_t>(std::distance(right_begin, right_end));
+
             auto new_memory = std::make_shared<PhysicalMemory>();
-            new_memory->insert(new_memory->end(), left.backing_block->begin() + left.offset,
-                               left.backing_block->begin() + left.offset + left.size);
-            new_memory->insert(new_memory->end(), right.backing_block->begin() + right.offset,
-                               right.backing_block->begin() + right.offset + right.size);
-            left.backing_block = new_memory;
+            new_memory->reserve(left_size + right_size);
+            new_memory->insert(new_memory->end(), left_begin, left_end);
+            new_memory->insert(new_memory->end(), right_begin, right_end);
+
+            left.backing_block = std::move(new_memory);
             left.offset = 0;
         }
 
@@ -965,7 +967,7 @@ ResultVal<std::size_t> VMManager::SizeOfAllocatedVMAsInRange(VAddr address,
 
     VAddr cur_addr = address;
     auto iter = FindVMA(cur_addr);
-    ASSERT_MSG(iter != vma_map.end(), "SizeOfAllocatedVMAsInRange iter != end");
+    ASSERT(iter != vma_map.end());
 
     while (true) {
         const auto& vma = iter->second;
@@ -986,7 +988,7 @@ ResultVal<std::size_t> VMManager::SizeOfAllocatedVMAsInRange(VAddr address,
         // Advance to the next block.
         cur_addr = vma_end;
         iter = std::next(iter);
-        ASSERT_MSG(iter != vma_map.end(), "SizeOfAllocatedVMAsInRange iter != end");
+        ASSERT(iter != vma_map.end());
     }
 
     return MakeResult(mapped_size);
@@ -1000,7 +1002,7 @@ ResultVal<std::size_t> VMManager::SizeOfUnmappablePhysicalMemoryInRange(VAddr ad
 
     VAddr cur_addr = address;
     auto iter = FindVMA(cur_addr);
-    ASSERT_MSG(iter != vma_map.end(), "SizeOfUnmappablePhysicalMemoryInRange iter != end");
+    ASSERT(iter != vma_map.end());
 
     while (true) {
         const auto& vma = iter->second;
@@ -1029,7 +1031,7 @@ ResultVal<std::size_t> VMManager::SizeOfUnmappablePhysicalMemoryInRange(VAddr ad
         // Advance to the next block.
         cur_addr = vma_end;
         iter = std::next(iter);
-        ASSERT_MSG(iter != vma_map.end(), "SizeOfUnmappablePhysicalMemoryInRange iter != end");
+        ASSERT(iter != vma_map.end());
     }
 
     return MakeResult(mapped_size);

src/core/hle/kernel/vm_manager.h

@@ -454,8 +454,8 @@ public:
 
     /// Maps memory at a given address.
     ///
-    /// @param addr The virtual address to map memory at.
-    /// @param size The amount of memory to map.
+    /// @param target The virtual address to map memory at.
+    /// @param size   The amount of memory to map.
     ///
     /// @note The destination address must lie within the Map region.
     ///
@@ -468,8 +468,8 @@ public:
 
     /// Unmaps memory at a given address.
     ///
-    /// @param addr The virtual address to unmap memory at.
-    /// @param size The amount of memory to unmap.
+    /// @param target The virtual address to unmap memory at.
+    /// @param size   The amount of memory to unmap.
     ///
     /// @note The destination address must lie within the Map region.
     ///

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

@@ -56,7 +56,8 @@ struct LaunchParameters {
 };
 static_assert(sizeof(LaunchParameters) == 0x88);
 
-IWindowController::IWindowController() : ServiceFramework("IWindowController") {
+IWindowController::IWindowController(Core::System& system_)
+    : ServiceFramework("IWindowController"), system{system_} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {0, nullptr, "CreateWindow"},
@@ -75,7 +76,7 @@ IWindowController::IWindowController() : ServiceFramework("IWindowController") {
 IWindowController::~IWindowController() = default;
 
 void IWindowController::GetAppletResourceUserId(Kernel::HLERequestContext& ctx) {
-    const u64 process_id = Core::System::GetInstance().Kernel().CurrentProcess()->GetProcessID();
+    const u64 process_id = system.CurrentProcess()->GetProcessID();
 
     LOG_DEBUG(Service_AM, "called. Process ID=0x{:016X}", process_id);
 
@@ -231,8 +232,9 @@ IDebugFunctions::IDebugFunctions() : ServiceFramework{"IDebugFunctions"} {
 
 IDebugFunctions::~IDebugFunctions() = default;
 
-ISelfController::ISelfController(std::shared_ptr<NVFlinger::NVFlinger> nvflinger)
-    : ServiceFramework("ISelfController"), nvflinger(std::move(nvflinger)) {
+ISelfController::ISelfController(Core::System& system_,
+                                 std::shared_ptr<NVFlinger::NVFlinger> nvflinger_)
+    : ServiceFramework("ISelfController"), nvflinger(std::move(nvflinger_)) {
     // clang-format off
     static const FunctionInfo functions[] = {
         {0, nullptr, "Exit"},
@@ -280,7 +282,7 @@ ISelfController::ISelfController(std::shared_ptr<NVFlinger::NVFlinger> nvflinger
 
     RegisterHandlers(functions);
 
-    auto& kernel = Core::System::GetInstance().Kernel();
+    auto& kernel = system_.Kernel();
     launchable_event = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Manual,
                                                               "ISelfController:LaunchableEvent");
 
@@ -501,8 +503,7 @@ void ISelfController::GetAccumulatedSuspendedTickChangedEvent(Kernel::HLERequest
     rb.PushCopyObjects(accumulated_suspended_tick_changed_event.readable);
 }
 
-AppletMessageQueue::AppletMessageQueue() {
-    auto& kernel = Core::System::GetInstance().Kernel();
+AppletMessageQueue::AppletMessageQueue(Kernel::KernelCore& kernel) {
     on_new_message = Kernel::WritableEvent::CreateEventPair(kernel, Kernel::ResetType::Manual,
                                                             "AMMessageQueue:OnMessageRecieved");
     on_operation_mode_changed = Kernel::WritableEvent::CreateEventPair(
@@ -937,9 +938,8 @@ void IStorageAccessor::Read(Kernel::HLERequestContext& ctx) {
     rb.Push(RESULT_SUCCESS);
 }
 
-ILibraryAppletCreator::ILibraryAppletCreator(u64 current_process_title_id)
-    : ServiceFramework("ILibraryAppletCreator"),
-      current_process_title_id(current_process_title_id) {
+ILibraryAppletCreator::ILibraryAppletCreator(Core::System& system_)
+    : ServiceFramework("ILibraryAppletCreator"), system{system_} {
     static const FunctionInfo functions[] = {
         {0, &ILibraryAppletCreator::CreateLibraryApplet, "CreateLibraryApplet"},
         {1, nullptr, "TerminateAllLibraryApplets"},
@@ -961,8 +961,8 @@ void ILibraryAppletCreator::CreateLibraryApplet(Kernel::HLERequestContext& ctx)
     LOG_DEBUG(Service_AM, "called with applet_id={:08X}, applet_mode={:08X}",
               static_cast<u32>(applet_id), applet_mode);
 
-    const auto& applet_manager{Core::System::GetInstance().GetAppletManager()};
-    const auto applet = applet_manager.GetApplet(applet_id, current_process_title_id);
+    const auto& applet_manager{system.GetAppletManager()};
+    const auto applet = applet_manager.GetApplet(applet_id);
 
     if (applet == nullptr) {
         LOG_ERROR(Service_AM, "Applet doesn't exist! applet_id={}", static_cast<u32>(applet_id));
@@ -999,8 +999,7 @@ void ILibraryAppletCreator::CreateTransferMemoryStorage(Kernel::HLERequestContex
     const auto handle{rp.Pop<Kernel::Handle>()};
 
     const auto transfer_mem =
-        Core::System::GetInstance().CurrentProcess()->GetHandleTable().Get<Kernel::TransferMemory>(
-            handle);
+        system.CurrentProcess()->GetHandleTable().Get<Kernel::TransferMemory>(handle);
 
     if (transfer_mem == nullptr) {
         LOG_ERROR(Service_AM, "shared_mem is a nullpr for handle={:08X}", handle);
@@ -1018,7 +1017,8 @@ void ILibraryAppletCreator::CreateTransferMemoryStorage(Kernel::HLERequestContex
     rb.PushIpcInterface(std::make_shared<IStorage>(std::move(memory)));
 }
 
-IApplicationFunctions::IApplicationFunctions() : ServiceFramework("IApplicationFunctions") {
+IApplicationFunctions::IApplicationFunctions(Core::System& system_)
+    : ServiceFramework("IApplicationFunctions"), system{system_} {
     // clang-format off
     static const FunctionInfo functions[] = {
         {1, &IApplicationFunctions::PopLaunchParameter, "PopLaunchParameter"},
@@ -1180,7 +1180,7 @@ void IApplicationFunctions::GetDesiredLanguage(Kernel::HLERequestContext& ctx) {
     // Get supported languages from NACP, if possible
     // Default to 0 (all languages supported)
     u32 supported_languages = 0;
-    FileSys::PatchManager pm{Core::System::GetInstance().CurrentProcess()->GetTitleID()};
+    FileSys::PatchManager pm{system.CurrentProcess()->GetTitleID()};
 
     const auto res = pm.GetControlMetadata();
     if (res.first != nullptr) {
@@ -1188,8 +1188,8 @@ void IApplicationFunctions::GetDesiredLanguage(Kernel::HLERequestContext& ctx) {
     }
 
     // Call IApplicationManagerInterface implementation.
-    auto& service_manager = Core::System::GetInstance().ServiceManager();
-    auto ns_am2 = service_manager.GetService<Service::NS::NS>("ns:am2");
+    auto& service_manager = system.ServiceManager();
+    auto ns_am2 = service_manager.GetService<NS::NS>("ns:am2");
     auto app_man = ns_am2->GetApplicationManagerInterface();
 
     // Get desired application language
@@ -1261,8 +1261,8 @@ void IApplicationFunctions::ExtendSaveData(Kernel::HLERequestContext& ctx) {
               "new_journal={:016X}",
               static_cast<u8>(type), user_id[1], user_id[0], new_normal_size, new_journal_size);
 
-    FileSystem::WriteSaveDataSize(type, Core::CurrentProcess()->GetTitleID(), user_id,
-                                  {new_normal_size, new_journal_size});
+    const auto title_id = system.CurrentProcess()->GetTitleID();
+    FileSystem::WriteSaveDataSize(type, title_id, user_id, {new_normal_size, new_journal_size});
 
     IPC::ResponseBuilder rb{ctx, 4};
     rb.Push(RESULT_SUCCESS);
@@ -1281,8 +1281,8 @@ void IApplicationFunctions::GetSaveDataSize(Kernel::HLERequestContext& ctx) {
     LOG_DEBUG(Service_AM, "called with type={:02X}, user_id={:016X}{:016X}", static_cast<u8>(type),
               user_id[1], user_id[0]);
 
-    const auto size =
-        FileSystem::ReadSaveDataSize(type, Core::CurrentProcess()->GetTitleID(), user_id);
+    const auto title_id = system.CurrentProcess()->GetTitleID();
+    const auto size = FileSystem::ReadSaveDataSize(type, title_id, user_id);
 
     IPC::ResponseBuilder rb{ctx, 6};
     rb.Push(RESULT_SUCCESS);
@@ -1300,9 +1300,9 @@ void IApplicationFunctions::GetGpuErrorDetectedSystemEvent(Kernel::HLERequestCon
 
 void InstallInterfaces(SM::ServiceManager& service_manager,
                        std::shared_ptr<NVFlinger::NVFlinger> nvflinger, Core::System& system) {
-    auto message_queue = std::make_shared<AppletMessageQueue>();
-    message_queue->PushMessage(AppletMessageQueue::AppletMessage::FocusStateChanged); // Needed on
-                                                                                      // game boot
+    auto message_queue = std::make_shared<AppletMessageQueue>(system.Kernel());
+    // Needed on game boot
+    message_queue->PushMessage(AppletMessageQueue::AppletMessage::FocusStateChanged);
 
     std::make_shared<AppletAE>(nvflinger, message_queue, system)->InstallAsService(service_manager);
     std::make_shared<AppletOE>(nvflinger, message_queue, system)->InstallAsService(service_manager);

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

@@ -10,12 +10,15 @@
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/service.h"
 
-namespace Service {
-namespace NVFlinger {
+namespace Kernel {
+class KernelCore;
+}
+
+namespace Service::NVFlinger {
 class NVFlinger;
 }
 
-namespace AM {
+namespace Service::AM {
 
 enum SystemLanguage {
     Japanese = 0,
@@ -47,7 +50,7 @@ public:
         PerformanceModeChanged = 31,
     };
 
-    AppletMessageQueue();
+    explicit AppletMessageQueue(Kernel::KernelCore& kernel);
     ~AppletMessageQueue();
 
     const Kernel::SharedPtr<Kernel::ReadableEvent>& GetMesssageRecieveEvent() const;
@@ -65,12 +68,14 @@ private:
 
 class IWindowController final : public ServiceFramework<IWindowController> {
 public:
-    IWindowController();
+    explicit IWindowController(Core::System& system_);
     ~IWindowController() override;
 
 private:
     void GetAppletResourceUserId(Kernel::HLERequestContext& ctx);
     void AcquireForegroundRights(Kernel::HLERequestContext& ctx);
+
+    Core::System& system;
 };
 
 class IAudioController final : public ServiceFramework<IAudioController> {
@@ -113,7 +118,8 @@ public:
 
 class ISelfController final : public ServiceFramework<ISelfController> {
 public:
-    explicit ISelfController(std::shared_ptr<NVFlinger::NVFlinger> nvflinger);
+    explicit ISelfController(Core::System& system_,
+                             std::shared_ptr<NVFlinger::NVFlinger> nvflinger_);
     ~ISelfController() override;
 
 private:
@@ -208,7 +214,7 @@ private:
 
 class ILibraryAppletCreator final : public ServiceFramework<ILibraryAppletCreator> {
 public:
-    ILibraryAppletCreator(u64 current_process_title_id);
+    explicit ILibraryAppletCreator(Core::System& system_);
     ~ILibraryAppletCreator() override;
 
 private:
@@ -216,12 +222,12 @@ private:
     void CreateStorage(Kernel::HLERequestContext& ctx);
     void CreateTransferMemoryStorage(Kernel::HLERequestContext& ctx);
 
-    u64 current_process_title_id;
+    Core::System& system;
 };
 
 class IApplicationFunctions final : public ServiceFramework<IApplicationFunctions> {
 public:
-    IApplicationFunctions();
+    explicit IApplicationFunctions(Core::System& system_);
     ~IApplicationFunctions() override;
 
 private:
@@ -245,6 +251,7 @@ private:
     void GetGpuErrorDetectedSystemEvent(Kernel::HLERequestContext& ctx);
 
     Kernel::EventPair gpu_error_detected_event;
+    Core::System& system;
 };
 
 class IHomeMenuFunctions final : public ServiceFramework<IHomeMenuFunctions> {
@@ -278,5 +285,4 @@ public:
 void InstallInterfaces(SM::ServiceManager& service_manager,
                        std::shared_ptr<NVFlinger::NVFlinger> nvflinger, Core::System& system);
 
-} // namespace AM
-} // namespace Service
+} // namespace Service::AM

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

@@ -50,7 +50,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<ISelfController>(nvflinger);
+        rb.PushIpcInterface<ISelfController>(system, nvflinger);
     }
 
     void GetWindowController(Kernel::HLERequestContext& ctx) {
@@ -58,7 +58,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<IWindowController>();
+        rb.PushIpcInterface<IWindowController>(system);
     }
 
     void GetAudioController(Kernel::HLERequestContext& ctx) {
@@ -98,7 +98,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<ILibraryAppletCreator>(system.CurrentProcess()->GetTitleID());
+        rb.PushIpcInterface<ILibraryAppletCreator>(system);
     }
 
     void GetApplicationFunctions(Kernel::HLERequestContext& ctx) {
@@ -106,7 +106,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<IApplicationFunctions>();
+        rb.PushIpcInterface<IApplicationFunctions>(system);
     }
 
     std::shared_ptr<NVFlinger::NVFlinger> nvflinger;
@@ -154,7 +154,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<ISelfController>(nvflinger);
+        rb.PushIpcInterface<ISelfController>(system, nvflinger);
     }
 
     void GetWindowController(Kernel::HLERequestContext& ctx) {
@@ -162,7 +162,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<IWindowController>();
+        rb.PushIpcInterface<IWindowController>(system);
     }
 
     void GetAudioController(Kernel::HLERequestContext& ctx) {
@@ -194,7 +194,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<ILibraryAppletCreator>(system.CurrentProcess()->GetTitleID());
+        rb.PushIpcInterface<ILibraryAppletCreator>(system);
     }
 
     void GetHomeMenuFunctions(Kernel::HLERequestContext& ctx) {

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

@@ -4,7 +4,6 @@
 
 #include "common/logging/log.h"
 #include "core/hle/ipc_helpers.h"
-#include "core/hle/kernel/process.h"
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applet_oe.h"
 #include "core/hle/service/nvflinger/nvflinger.h"
@@ -64,7 +63,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<IWindowController>();
+        rb.PushIpcInterface<IWindowController>(system);
     }
 
     void GetSelfController(Kernel::HLERequestContext& ctx) {
@@ -72,7 +71,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<ISelfController>(nvflinger);
+        rb.PushIpcInterface<ISelfController>(system, nvflinger);
     }
 
     void GetCommonStateGetter(Kernel::HLERequestContext& ctx) {
@@ -88,7 +87,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<ILibraryAppletCreator>(system.CurrentProcess()->GetTitleID());
+        rb.PushIpcInterface<ILibraryAppletCreator>(system);
     }
 
     void GetApplicationFunctions(Kernel::HLERequestContext& ctx) {
@@ -96,7 +95,7 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 2, 0, 1};
         rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<IApplicationFunctions>();
+        rb.PushIpcInterface<IApplicationFunctions>(system);
     }
 
     std::shared_ptr<NVFlinger::NVFlinger> nvflinger;

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

@@ -23,8 +23,7 @@
 
 namespace Service::AM::Applets {
 
-AppletDataBroker::AppletDataBroker() {
-    auto& kernel = Core::System::GetInstance().Kernel();
+AppletDataBroker::AppletDataBroker(Kernel::KernelCore& kernel) {
     state_changed_event = Kernel::WritableEvent::CreateEventPair(
         kernel, Kernel::ResetType::Manual, "ILibraryAppletAccessor:StateChangedEvent");
     pop_out_data_event = Kernel::WritableEvent::CreateEventPair(
@@ -121,7 +120,7 @@ Kernel::SharedPtr<Kernel::ReadableEvent> AppletDataBroker::GetStateChangedEvent(
     return state_changed_event.readable;
 }
 
-Applet::Applet() = default;
+Applet::Applet(Kernel::KernelCore& kernel_) : broker{kernel_} {}
 
 Applet::~Applet() = default;
 
@@ -154,7 +153,7 @@ AppletFrontendSet::AppletFrontendSet(AppletFrontendSet&&) noexcept = default;
 
 AppletFrontendSet& AppletFrontendSet::operator=(AppletFrontendSet&&) noexcept = default;
 
-AppletManager::AppletManager() = default;
+AppletManager::AppletManager(Core::System& system_) : system{system_} {}
 
 AppletManager::~AppletManager() = default;
 
@@ -216,28 +215,28 @@ void AppletManager::ClearAll() {
     frontend = {};
 }
 
-std::shared_ptr<Applet> AppletManager::GetApplet(AppletId id, u64 current_process_title_id) const {
+std::shared_ptr<Applet> AppletManager::GetApplet(AppletId id) const {
     switch (id) {
     case AppletId::Auth:
-        return std::make_shared<Auth>(*frontend.parental_controls);
+        return std::make_shared<Auth>(system, *frontend.parental_controls);
     case AppletId::Error:
-        return std::make_shared<Error>(*frontend.error);
+        return std::make_shared<Error>(system, *frontend.error);
     case AppletId::ProfileSelect:
-        return std::make_shared<ProfileSelect>(*frontend.profile_select);
+        return std::make_shared<ProfileSelect>(system, *frontend.profile_select);
     case AppletId::SoftwareKeyboard:
-        return std::make_shared<SoftwareKeyboard>(*frontend.software_keyboard);
+        return std::make_shared<SoftwareKeyboard>(system, *frontend.software_keyboard);
     case AppletId::PhotoViewer:
-        return std::make_shared<PhotoViewer>(*frontend.photo_viewer);
+        return std::make_shared<PhotoViewer>(system, *frontend.photo_viewer);
     case AppletId::LibAppletShop:
-        return std::make_shared<WebBrowser>(*frontend.web_browser, current_process_title_id,
+        return std::make_shared<WebBrowser>(system, *frontend.web_browser,
                                             frontend.e_commerce.get());
     case AppletId::LibAppletOff:
-        return std::make_shared<WebBrowser>(*frontend.web_browser, current_process_title_id);
+        return std::make_shared<WebBrowser>(system, *frontend.web_browser);
     default:
         UNIMPLEMENTED_MSG(
             "No backend implementation exists for applet_id={:02X}! Falling back to stub applet.",
             static_cast<u8>(id));
-        return std::make_shared<StubApplet>(id);
+        return std::make_shared<StubApplet>(system, id);
     }
 }
 

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

@@ -12,6 +12,10 @@
 
 union ResultCode;
 
+namespace Core {
+class System;
+}
+
 namespace Core::Frontend {
 class ECommerceApplet;
 class ErrorApplet;
@@ -22,6 +26,10 @@ class SoftwareKeyboardApplet;
 class WebBrowserApplet;
 } // namespace Core::Frontend
 
+namespace Kernel {
+class KernelCore;
+}
+
 namespace Service::AM {
 
 class IStorage;
@@ -53,7 +61,7 @@ enum class AppletId : u32 {
 
 class AppletDataBroker final {
 public:
-    AppletDataBroker();
+    explicit AppletDataBroker(Kernel::KernelCore& kernel_);
     ~AppletDataBroker();
 
     struct RawChannelData {
@@ -108,7 +116,7 @@ private:
 
 class Applet {
 public:
-    Applet();
+    explicit Applet(Kernel::KernelCore& kernel_);
     virtual ~Applet();
 
     virtual void Initialize();
@@ -179,7 +187,7 @@ struct AppletFrontendSet {
 
 class AppletManager {
 public:
-    AppletManager();
+    explicit AppletManager(Core::System& system_);
     ~AppletManager();
 
     void SetAppletFrontendSet(AppletFrontendSet set);
@@ -187,10 +195,11 @@ public:
     void SetDefaultAppletsIfMissing();
     void ClearAll();
 
-    std::shared_ptr<Applet> GetApplet(AppletId id, u64 current_process_title_id) const;
+    std::shared_ptr<Applet> GetApplet(AppletId id) const;
 
 private:
     AppletFrontendSet frontend;
+    Core::System& system;
 };
 
 } // namespace Applets

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

@@ -85,7 +85,8 @@ ResultCode Decode64BitError(u64 error) {
 
 } // Anonymous namespace
 
-Error::Error(const Core::Frontend::ErrorApplet& frontend) : frontend(frontend) {}
+Error::Error(Core::System& system_, const Core::Frontend::ErrorApplet& frontend_)
+    : Applet{system_.Kernel()}, frontend(frontend_), system{system_} {}
 
 Error::~Error() = default;
 
@@ -145,8 +146,8 @@ void Error::Execute() {
     }
 
     const auto callback = [this] { DisplayCompleted(); };
-    const auto title_id = Core::CurrentProcess()->GetTitleID();
-    const auto& reporter{Core::System::GetInstance().GetReporter()};
+    const auto title_id = system.CurrentProcess()->GetTitleID();
+    const auto& reporter{system.GetReporter()};
 
     switch (mode) {
     case ErrorAppletMode::ShowError:

src/core/hle/service/am/applets/error.h

@@ -7,6 +7,10 @@
 #include "core/hle/result.h"
 #include "core/hle/service/am/applets/applets.h"
 
+namespace Core {
+class System;
+}
+
 namespace Service::AM::Applets {
 
 enum class ErrorAppletMode : u8 {
@@ -21,7 +25,7 @@ enum class ErrorAppletMode : u8 {
 
 class Error final : public Applet {
 public:
-    explicit Error(const Core::Frontend::ErrorApplet& frontend);
+    explicit Error(Core::System& system_, const Core::Frontend::ErrorApplet& frontend_);
     ~Error() override;
 
     void Initialize() override;
@@ -42,6 +46,7 @@ private:
     std::unique_ptr<ErrorArguments> args;
 
     bool complete = false;
+    Core::System& system;
 };
 
 } // namespace Service::AM::Applets

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

@@ -37,7 +37,8 @@ static void LogCurrentStorage(AppletDataBroker& broker, std::string_view prefix)
     }
 }
 
-Auth::Auth(Core::Frontend::ParentalControlsApplet& frontend) : frontend(frontend) {}
+Auth::Auth(Core::System& system_, Core::Frontend::ParentalControlsApplet& frontend_)
+    : Applet{system_.Kernel()}, frontend(frontend_) {}
 
 Auth::~Auth() = default;
 
@@ -151,7 +152,8 @@ void Auth::AuthFinished(bool successful) {
     broker.SignalStateChanged();
 }
 
-PhotoViewer::PhotoViewer(const Core::Frontend::PhotoViewerApplet& frontend) : frontend(frontend) {}
+PhotoViewer::PhotoViewer(Core::System& system_, const Core::Frontend::PhotoViewerApplet& frontend_)
+    : Applet{system_.Kernel()}, frontend(frontend_), system{system_} {}
 
 PhotoViewer::~PhotoViewer() = default;
 
@@ -185,7 +187,7 @@ void PhotoViewer::Execute() {
     const auto callback = [this] { ViewFinished(); };
     switch (mode) {
     case PhotoViewerAppletMode::CurrentApp:
-        frontend.ShowPhotosForApplication(Core::CurrentProcess()->GetTitleID(), callback);
+        frontend.ShowPhotosForApplication(system.CurrentProcess()->GetTitleID(), callback);
         break;
     case PhotoViewerAppletMode::AllApps:
         frontend.ShowAllPhotos(callback);
@@ -200,7 +202,8 @@ void PhotoViewer::ViewFinished() {
     broker.SignalStateChanged();
 }
 
-StubApplet::StubApplet(AppletId id) : id(id) {}
+StubApplet::StubApplet(Core::System& system_, AppletId id_)
+    : Applet{system_.Kernel()}, id(id_), system{system_} {}
 
 StubApplet::~StubApplet() = default;
 
@@ -209,7 +212,7 @@ void StubApplet::Initialize() {
     Applet::Initialize();
 
     const auto data = broker.PeekDataToAppletForDebug();
-    Core::System::GetInstance().GetReporter().SaveUnimplementedAppletReport(
+    system.GetReporter().SaveUnimplementedAppletReport(
         static_cast<u32>(id), common_args.arguments_version, common_args.library_version,
         common_args.theme_color, common_args.play_startup_sound, common_args.system_tick,
         data.normal, data.interactive);

src/core/hle/service/am/applets/general_backend.h

@@ -6,6 +6,10 @@
 
 #include "core/hle/service/am/applets/applets.h"
 
+namespace Core {
+class System;
+}
+
 namespace Service::AM::Applets {
 
 enum class AuthAppletType : u32 {
@@ -16,7 +20,7 @@ enum class AuthAppletType : u32 {
 
 class Auth final : public Applet {
 public:
-    explicit Auth(Core::Frontend::ParentalControlsApplet& frontend);
+    explicit Auth(Core::System& system_, Core::Frontend::ParentalControlsApplet& frontend_);
     ~Auth() override;
 
     void Initialize() override;
@@ -45,7 +49,7 @@ enum class PhotoViewerAppletMode : u8 {
 
 class PhotoViewer final : public Applet {
 public:
-    explicit PhotoViewer(const Core::Frontend::PhotoViewerApplet& frontend);
+    explicit PhotoViewer(Core::System& system_, const Core::Frontend::PhotoViewerApplet& frontend_);
     ~PhotoViewer() override;
 
     void Initialize() override;
@@ -60,11 +64,12 @@ private:
     const Core::Frontend::PhotoViewerApplet& frontend;
     bool complete = false;
     PhotoViewerAppletMode mode = PhotoViewerAppletMode::CurrentApp;
+    Core::System& system;
 };
 
 class StubApplet final : public Applet {
 public:
-    explicit StubApplet(AppletId id);
+    explicit StubApplet(Core::System& system_, AppletId id_);
     ~StubApplet() override;
 
     void Initialize() override;
@@ -76,6 +81,7 @@ public:
 
 private:
     AppletId id;
+    Core::System& system;
 };
 
 } // namespace Service::AM::Applets

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

@@ -15,8 +15,9 @@ namespace Service::AM::Applets {
 
 constexpr ResultCode ERR_USER_CANCELLED_SELECTION{ErrorModule::Account, 1};
 
-ProfileSelect::ProfileSelect(const Core::Frontend::ProfileSelectApplet& frontend)
-    : frontend(frontend) {}
+ProfileSelect::ProfileSelect(Core::System& system_,
+                             const Core::Frontend::ProfileSelectApplet& frontend_)
+    : Applet{system_.Kernel()}, frontend(frontend_) {}
 
 ProfileSelect::~ProfileSelect() = default;
 

src/core/hle/service/am/applets/profile_select.h

@@ -11,6 +11,10 @@
 #include "core/hle/result.h"
 #include "core/hle/service/am/applets/applets.h"
 
+namespace Core {
+class System;
+}
+
 namespace Service::AM::Applets {
 
 struct UserSelectionConfig {
@@ -29,7 +33,8 @@ static_assert(sizeof(UserSelectionOutput) == 0x18, "UserSelectionOutput has inco
 
 class ProfileSelect final : public Applet {
 public:
-    explicit ProfileSelect(const Core::Frontend::ProfileSelectApplet& frontend);
+    explicit ProfileSelect(Core::System& system_,
+                           const Core::Frontend::ProfileSelectApplet& frontend_);
     ~ProfileSelect() override;
 
     void Initialize() override;

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

@@ -39,8 +39,9 @@ static Core::Frontend::SoftwareKeyboardParameters ConvertToFrontendParameters(
     return params;
 }
 
-SoftwareKeyboard::SoftwareKeyboard(const Core::Frontend::SoftwareKeyboardApplet& frontend)
-    : frontend(frontend) {}
+SoftwareKeyboard::SoftwareKeyboard(Core::System& system_,
+                                   const Core::Frontend::SoftwareKeyboardApplet& frontend_)
+    : Applet{system_.Kernel()}, frontend(frontend_) {}
 
 SoftwareKeyboard::~SoftwareKeyboard() = default;
 

src/core/hle/service/am/applets/software_keyboard.h

@@ -16,6 +16,10 @@
 
 union ResultCode;
 
+namespace Core {
+class System;
+}
+
 namespace Service::AM::Applets {
 
 enum class KeysetDisable : u32 {
@@ -55,7 +59,8 @@ static_assert(sizeof(KeyboardConfig) == 0x3E0, "KeyboardConfig has incorrect siz
 
 class SoftwareKeyboard final : public Applet {
 public:
-    explicit SoftwareKeyboard(const Core::Frontend::SoftwareKeyboardApplet& frontend);
+    explicit SoftwareKeyboard(Core::System& system_,
+                              const Core::Frontend::SoftwareKeyboardApplet& frontend_);
     ~SoftwareKeyboard() override;
 
     void Initialize() override;

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

@@ -190,8 +190,9 @@ std::map<WebArgTLVType, std::vector<u8>> GetWebArguments(const std::vector<u8>&
     return out;
 }
 
-FileSys::VirtualFile GetApplicationRomFS(u64 title_id, FileSys::ContentRecordType type) {
-    const auto& installed{Core::System::GetInstance().GetContentProvider()};
+FileSys::VirtualFile GetApplicationRomFS(const Core::System& system, u64 title_id,
+                                         FileSys::ContentRecordType type) {
+    const auto& installed{system.GetContentProvider()};
     const auto res = installed.GetEntry(title_id, type);
 
     if (res != nullptr) {
@@ -207,10 +208,10 @@ FileSys::VirtualFile GetApplicationRomFS(u64 title_id, FileSys::ContentRecordTyp
 
 } // Anonymous namespace
 
-WebBrowser::WebBrowser(Core::Frontend::WebBrowserApplet& frontend, u64 current_process_title_id,
-                       Core::Frontend::ECommerceApplet* frontend_e_commerce)
-    : frontend(frontend), frontend_e_commerce(frontend_e_commerce),
-      current_process_title_id(current_process_title_id) {}
+WebBrowser::WebBrowser(Core::System& system_, Core::Frontend::WebBrowserApplet& frontend_,
+                       Core::Frontend::ECommerceApplet* frontend_e_commerce_)
+    : Applet{system_.Kernel()}, frontend(frontend_),
+      frontend_e_commerce(frontend_e_commerce_), system{system_} {}
 
 WebBrowser::~WebBrowser() = default;
 
@@ -266,7 +267,7 @@ void WebBrowser::UnpackRomFS() {
     ASSERT(offline_romfs != nullptr);
     const auto dir =
         FileSys::ExtractRomFS(offline_romfs, FileSys::RomFSExtractionType::SingleDiscard);
-    const auto& vfs{Core::System::GetInstance().GetFilesystem()};
+    const auto& vfs{system.GetFilesystem()};
     const auto temp_dir = vfs->CreateDirectory(temporary_dir, FileSys::Mode::ReadWrite);
     FileSys::VfsRawCopyD(dir, temp_dir);
 
@@ -470,10 +471,10 @@ void WebBrowser::InitializeOffline() {
     }
 
     if (title_id == 0) {
-        title_id = current_process_title_id;
+        title_id = system.CurrentProcess()->GetTitleID();
     }
 
-    offline_romfs = GetApplicationRomFS(title_id, type);
+    offline_romfs = GetApplicationRomFS(system, title_id, type);
     if (offline_romfs == nullptr) {
         status = ResultCode(-1);
         LOG_ERROR(Service_AM, "Failed to find offline data for request!");

src/core/hle/service/am/applets/web_browser.h

@@ -9,6 +9,10 @@
 #include "core/hle/service/am/am.h"
 #include "core/hle/service/am/applets/applets.h"
 
+namespace Core {
+class System;
+}
+
 namespace Service::AM::Applets {
 
 enum class ShimKind : u32;
@@ -17,8 +21,8 @@ enum class WebArgTLVType : u16;
 
 class WebBrowser final : public Applet {
 public:
-    WebBrowser(Core::Frontend::WebBrowserApplet& frontend, u64 current_process_title_id,
-               Core::Frontend::ECommerceApplet* frontend_e_commerce = nullptr);
+    WebBrowser(Core::System& system_, Core::Frontend::WebBrowserApplet& frontend_,
+               Core::Frontend::ECommerceApplet* frontend_e_commerce_ = nullptr);
 
     ~WebBrowser() override;
 
@@ -59,8 +63,6 @@ private:
     bool unpacked = false;
     ResultCode status = RESULT_SUCCESS;
 
-    u64 current_process_title_id;
-
     ShimKind kind;
     std::map<WebArgTLVType, std::vector<u8>> args;
 
@@ -74,6 +76,8 @@ private:
     std::optional<u128> user_id;
     std::optional<bool> shop_full_display;
     std::string shop_extra_parameter;
+
+    Core::System& system;
 };
 
 } // namespace Service::AM::Applets

src/core/loader/nro.cpp

@@ -258,6 +258,15 @@ ResultStatus AppLoader_NRO::ReadTitle(std::string& title) {
     return ResultStatus::Success;
 }
 
+ResultStatus AppLoader_NRO::ReadControlData(FileSys::NACP& control) {
+    if (nacp == nullptr) {
+        return ResultStatus::ErrorNoControl;
+    }
+
+    control = *nacp;
+    return ResultStatus::Success;
+}
+
 bool AppLoader_NRO::IsRomFSUpdatable() const {
     return false;
 }

src/core/loader/nro.h

@@ -43,6 +43,7 @@ public:
     ResultStatus ReadProgramId(u64& out_program_id) override;
     ResultStatus ReadRomFS(FileSys::VirtualFile& dir) override;
     ResultStatus ReadTitle(std::string& title) override;
+    ResultStatus ReadControlData(FileSys::NACP& control) override;
     bool IsRomFSUpdatable() const override;
 
 private:

src/video_core/engines/kepler_compute.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <bitset>
 #include "common/assert.h"
 #include "common/logging/log.h"
 #include "core/core.h"
@@ -49,6 +50,33 @@ void KeplerCompute::CallMethod(const GPU::MethodCall& method_call) {
     }
 }
 
+Tegra::Texture::FullTextureInfo KeplerCompute::GetTexture(std::size_t offset) const {
+    const std::bitset<8> cbuf_mask = launch_description.const_buffer_enable_mask.Value();
+    ASSERT(cbuf_mask[regs.tex_cb_index]);
+
+    const auto& texinfo = launch_description.const_buffer_config[regs.tex_cb_index];
+    ASSERT(texinfo.Address() != 0);
+
+    const GPUVAddr address = texinfo.Address() + offset * sizeof(Texture::TextureHandle);
+    ASSERT(address < texinfo.Address() + texinfo.size);
+
+    const Texture::TextureHandle tex_handle{memory_manager.Read<u32>(address)};
+    return GetTextureInfo(tex_handle, offset);
+}
+
+Texture::FullTextureInfo KeplerCompute::GetTextureInfo(const Texture::TextureHandle tex_handle,
+                                                       std::size_t offset) const {
+    return Texture::FullTextureInfo{static_cast<u32>(offset), GetTICEntry(tex_handle.tic_id),
+                                    GetTSCEntry(tex_handle.tsc_id)};
+}
+
+u32 KeplerCompute::AccessConstBuffer32(u64 const_buffer, u64 offset) const {
+    const auto& buffer = launch_description.const_buffer_config[const_buffer];
+    u32 result;
+    std::memcpy(&result, memory_manager.GetPointer(buffer.Address() + offset), sizeof(u32));
+    return result;
+}
+
 void KeplerCompute::ProcessLaunch() {
     const GPUVAddr launch_desc_loc = regs.launch_desc_loc.Address();
     memory_manager.ReadBlockUnsafe(launch_desc_loc, &launch_description,
@@ -60,4 +88,29 @@ void KeplerCompute::ProcessLaunch() {
     rasterizer.DispatchCompute(code_addr);
 }
 
+Texture::TICEntry KeplerCompute::GetTICEntry(u32 tic_index) const {
+    const GPUVAddr tic_address_gpu{regs.tic.Address() + tic_index * sizeof(Texture::TICEntry)};
+
+    Texture::TICEntry tic_entry;
+    memory_manager.ReadBlockUnsafe(tic_address_gpu, &tic_entry, sizeof(Texture::TICEntry));
+
+    const auto r_type{tic_entry.r_type.Value()};
+    const auto g_type{tic_entry.g_type.Value()};
+    const auto b_type{tic_entry.b_type.Value()};
+    const auto a_type{tic_entry.a_type.Value()};
+
+    // TODO(Subv): Different data types for separate components are not supported
+    DEBUG_ASSERT(r_type == g_type && r_type == b_type && r_type == a_type);
+
+    return tic_entry;
+}
+
+Texture::TSCEntry KeplerCompute::GetTSCEntry(u32 tsc_index) const {
+    const GPUVAddr tsc_address_gpu{regs.tsc.Address() + tsc_index * sizeof(Texture::TSCEntry)};
+
+    Texture::TSCEntry tsc_entry;
+    memory_manager.ReadBlockUnsafe(tsc_address_gpu, &tsc_entry, sizeof(Texture::TSCEntry));
+    return tsc_entry;
+}
+
 } // namespace Tegra::Engines

src/video_core/engines/kepler_compute.h

@@ -12,6 +12,7 @@
 #include "common/common_types.h"
 #include "video_core/engines/engine_upload.h"
 #include "video_core/gpu.h"
+#include "video_core/textures/texture.h"
 
 namespace Core {
 class System;
@@ -111,7 +112,7 @@ public:
 
                 INSERT_PADDING_WORDS(0x3FE);
 
-                u32 texture_const_buffer_index;
+                u32 tex_cb_index;
 
                 INSERT_PADDING_WORDS(0x374);
             };
@@ -149,7 +150,7 @@ public:
         union {
             BitField<0, 8, u32> const_buffer_enable_mask;
             BitField<29, 2, u32> cache_layout;
-        } memory_config;
+        };
 
         INSERT_PADDING_WORDS(0x8);
 
@@ -194,6 +195,14 @@ public:
     /// Write the value to the register identified by method.
     void CallMethod(const GPU::MethodCall& method_call);
 
+    Tegra::Texture::FullTextureInfo GetTexture(std::size_t offset) const;
+
+    /// Given a Texture Handle, returns the TSC and TIC entries.
+    Texture::FullTextureInfo GetTextureInfo(const Texture::TextureHandle tex_handle,
+                                            std::size_t offset) const;
+
+    u32 AccessConstBuffer32(u64 const_buffer, u64 offset) const;
+
 private:
     Core::System& system;
     VideoCore::RasterizerInterface& rasterizer;
@@ -201,6 +210,12 @@ private:
     Upload::State upload_state;
 
     void ProcessLaunch();
+
+    /// Retrieves information about a specific TIC entry from the TIC buffer.
+    Texture::TICEntry GetTICEntry(u32 tic_index) const;
+
+    /// Retrieves information about a specific TSC entry from the TSC buffer.
+    Texture::TSCEntry GetTSCEntry(u32 tsc_index) const;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
@@ -218,12 +233,12 @@ ASSERT_REG_POSITION(launch, 0xAF);
 ASSERT_REG_POSITION(tsc, 0x557);
 ASSERT_REG_POSITION(tic, 0x55D);
 ASSERT_REG_POSITION(code_loc, 0x582);
-ASSERT_REG_POSITION(texture_const_buffer_index, 0x982);
+ASSERT_REG_POSITION(tex_cb_index, 0x982);
 ASSERT_LAUNCH_PARAM_POSITION(program_start, 0x8);
 ASSERT_LAUNCH_PARAM_POSITION(grid_dim_x, 0xC);
 ASSERT_LAUNCH_PARAM_POSITION(shared_alloc, 0x11);
 ASSERT_LAUNCH_PARAM_POSITION(block_dim_x, 0x12);
-ASSERT_LAUNCH_PARAM_POSITION(memory_config, 0x14);
+ASSERT_LAUNCH_PARAM_POSITION(const_buffer_enable_mask, 0x14);
 ASSERT_LAUNCH_PARAM_POSITION(const_buffer_config, 0x1D);
 
 #undef ASSERT_REG_POSITION

src/video_core/engines/maxwell_3d.cpp

@@ -89,6 +89,9 @@ void Maxwell3D::InitializeRegisterDefaults() {
 
     // Commercial games seem to assume this value is enabled and nouveau sets this value manually.
     regs.rt_separate_frag_data = 1;
+
+    // Some games (like Super Mario Odyssey) assume that SRGB is enabled.
+    regs.framebuffer_srgb = 1;
 }
 
 #define DIRTY_REGS_POS(field_name) (offsetof(Maxwell3D::DirtyRegs, field_name))
@@ -244,7 +247,7 @@ void Maxwell3D::InitDirtySettings() {
     dirty_pointers[MAXWELL3D_REG_INDEX(polygon_offset_clamp)] = polygon_offset_dirty_reg;
 }
 
-void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {
+void Maxwell3D::CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters) {
     // Reset the current macro.
     executing_macro = 0;
 
@@ -252,7 +255,7 @@ void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {
     const u32 entry = ((method - MacroRegistersStart) >> 1) % macro_positions.size();
 
     // Execute the current macro.
-    macro_interpreter.Execute(macro_positions[entry], std::move(parameters));
+    macro_interpreter.Execute(macro_positions[entry], num_parameters, parameters);
 }
 
 void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
@@ -289,7 +292,8 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
 
         // Call the macro when there are no more parameters in the command buffer
         if (method_call.IsLastCall()) {
-            CallMacroMethod(executing_macro, std::move(macro_params));
+            CallMacroMethod(executing_macro, macro_params.size(), macro_params.data());
+            macro_params.clear();
         }
         return;
     }

src/video_core/engines/maxwell_3d.h

@@ -62,6 +62,7 @@ public:
         static constexpr std::size_t NumVertexAttributes = 32;
         static constexpr std::size_t NumVaryings = 31;
         static constexpr std::size_t NumTextureSamplers = 32;
+        static constexpr std::size_t NumImages = 8; // TODO(Rodrigo): Investigate this number
         static constexpr std::size_t NumClipDistances = 8;
         static constexpr std::size_t MaxShaderProgram = 6;
         static constexpr std::size_t MaxShaderStage = 5;
@@ -1307,9 +1308,10 @@ private:
     /**
      * Call a macro on this engine.
      * @param method Method to call
+     * @param num_parameters Number of arguments
      * @param parameters Arguments to the method call
      */
-    void CallMacroMethod(u32 method, std::vector<u32> parameters);
+    void CallMacroMethod(u32 method, std::size_t num_parameters, const u32* parameters);
 
     /// Handles writes to the macro uploading register.
     void ProcessMacroUpload(u32 data);

src/video_core/engines/shader_bytecode.h

@@ -674,6 +674,10 @@ union Instruction {
         BitField<48, 1, u64> is_signed;
     } shift;
 
+    union {
+        BitField<39, 1, u64> wrap;
+    } shr;
+
     union {
         BitField<39, 5, u64> shift_amount;
         BitField<48, 1, u64> negate_b;

src/video_core/macro_interpreter.cpp

@@ -14,11 +14,18 @@ namespace Tegra {
 
 MacroInterpreter::MacroInterpreter(Engines::Maxwell3D& maxwell3d) : maxwell3d(maxwell3d) {}
 
-void MacroInterpreter::Execute(u32 offset, std::vector<u32> parameters) {
+void MacroInterpreter::Execute(u32 offset, std::size_t num_parameters, const u32* parameters) {
     MICROPROFILE_SCOPE(MacroInterp);
     Reset();
+
     registers[1] = parameters[0];
-    this->parameters = std::move(parameters);
+
+    if (num_parameters > parameters_capacity) {
+        parameters_capacity = num_parameters;
+        this->parameters = std::make_unique<u32[]>(num_parameters);
+    }
+    std::memcpy(this->parameters.get(), parameters, num_parameters * sizeof(u32));
+    this->num_parameters = num_parameters;
 
     // Execute the code until we hit an exit condition.
     bool keep_executing = true;
@@ -27,7 +34,7 @@ void MacroInterpreter::Execute(u32 offset, std::vector<u32> parameters) {
     }
 
     // Assert the the macro used all the input parameters
-    ASSERT(next_parameter_index == this->parameters.size());
+    ASSERT(next_parameter_index == num_parameters);
 }
 
 void MacroInterpreter::Reset() {
@@ -35,7 +42,7 @@ void MacroInterpreter::Reset() {
     pc = 0;
     delayed_pc = {};
     method_address.raw = 0;
-    parameters.clear();
+    num_parameters = 0;
     // The next parameter index starts at 1, because $r1 already has the value of the first
     // parameter.
     next_parameter_index = 1;
@@ -229,7 +236,8 @@ void MacroInterpreter::ProcessResult(ResultOperation operation, u32 reg, u32 res
 }
 
 u32 MacroInterpreter::FetchParameter() {
-    return parameters.at(next_parameter_index++);
+    ASSERT(next_parameter_index < num_parameters);
+    return parameters[next_parameter_index++];
 }
 
 u32 MacroInterpreter::GetRegister(u32 register_id) const {

src/video_core/macro_interpreter.h

@@ -25,7 +25,7 @@ public:
      * @param offset Offset to start execution at.
      * @param parameters The parameters of the macro.
      */
-    void Execute(u32 offset, std::vector<u32> parameters);
+    void Execute(u32 offset, std::size_t num_parameters, const u32* parameters);
 
 private:
     enum class Operation : u32 {
@@ -162,10 +162,12 @@ private:
     MethodAddress method_address = {};
 
     /// Input parameters of the current macro.
-    std::vector<u32> parameters;
+    std::unique_ptr<u32[]> parameters;
+    std::size_t num_parameters = 0;
+    std::size_t parameters_capacity = 0;
     /// Index of the next parameter that will be fetched by the 'parm' instruction.
     u32 next_parameter_index = 0;
 
-    bool carry_flag{};
+    bool carry_flag = false;
 };
 } // namespace Tegra

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -331,7 +331,7 @@ void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
         const auto stage_enum = static_cast<Maxwell::ShaderStage>(stage);
         SetupDrawConstBuffers(stage_enum, shader);
         SetupDrawGlobalMemory(stage_enum, shader);
-        const auto texture_buffer_usage{SetupTextures(stage_enum, shader, base_bindings)};
+        const auto texture_buffer_usage{SetupDrawTextures(stage_enum, shader, base_bindings)};
 
         const ProgramVariant variant{base_bindings, primitive_mode, texture_buffer_usage};
         const auto [program_handle, next_bindings] = shader->GetProgramHandle(variant);
@@ -489,9 +489,6 @@ std::pair<bool, bool> RasterizerOpenGL::ConfigureFramebuffers(
                 // Assume that a surface will be written to if it is used as a framebuffer, even if
                 // the shader doesn't actually write to it.
                 texture_cache.MarkColorBufferInUse(*single_color_target);
-                // Workaround for and issue in nvidia drivers
-                // https://devtalk.nvidia.com/default/topic/776591/opengl/gl_framebuffer_srgb-functions-incorrectly/
-                state.framebuffer_srgb.enabled |= color_surface->GetSurfaceParams().srgb_conversion;
             }
 
             fbkey.is_single_buffer = true;
@@ -512,11 +509,6 @@ std::pair<bool, bool> RasterizerOpenGL::ConfigureFramebuffers(
                     // Assume that a surface will be written to if it is used as a framebuffer, even
                     // if the shader doesn't actually write to it.
                     texture_cache.MarkColorBufferInUse(index);
-                    // Enable sRGB only for supported formats
-                    // Workaround for and issue in nvidia drivers
-                    // https://devtalk.nvidia.com/default/topic/776591/opengl/gl_framebuffer_srgb-functions-incorrectly/
-                    state.framebuffer_srgb.enabled |=
-                        color_surface->GetSurfaceParams().srgb_conversion;
                 }
 
                 fbkey.color_attachments[index] =
@@ -537,8 +529,7 @@ std::pair<bool, bool> RasterizerOpenGL::ConfigureFramebuffers(
         texture_cache.MarkDepthBufferInUse();
 
         fbkey.zeta = depth_surface;
-        fbkey.stencil_enable = regs.stencil_enable &&
-                               depth_surface->GetSurfaceParams().type == SurfaceType::DepthStencil;
+        fbkey.stencil_enable = depth_surface->GetSurfaceParams().type == SurfaceType::DepthStencil;
     }
 
     texture_cache.GuardRenderTargets(false);
@@ -577,16 +568,15 @@ void RasterizerOpenGL::ConfigureClearFramebuffer(OpenGLState& current_state, boo
     if (depth_surface) {
         const auto& params = depth_surface->GetSurfaceParams();
         switch (params.type) {
-        case VideoCore::Surface::SurfaceType::Depth: {
+        case VideoCore::Surface::SurfaceType::Depth:
             depth_surface->Attach(GL_DEPTH_ATTACHMENT, GL_DRAW_FRAMEBUFFER);
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
             break;
-        }
-        case VideoCore::Surface::SurfaceType::DepthStencil: {
-            depth_surface->Attach(GL_DEPTH_ATTACHMENT, GL_DRAW_FRAMEBUFFER);
+        case VideoCore::Surface::SurfaceType::DepthStencil:
+            depth_surface->Attach(GL_DEPTH_STENCIL_ATTACHMENT, GL_DRAW_FRAMEBUFFER);
             break;
-        }
-        default: { UNIMPLEMENTED(); }
+        default:
+            UNIMPLEMENTED();
         }
     } else {
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
@@ -639,6 +629,7 @@ void RasterizerOpenGL::Clear() {
         ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear stencil but buffer is not enabled!");
         use_stencil = true;
         clear_state.stencil.test_enabled = true;
+
         if (regs.clear_flags.stencil) {
             // Stencil affects the clear so fill it with the used masks
             clear_state.stencil.front.test_func = GL_ALWAYS;
@@ -802,7 +793,11 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     }
 
     auto kernel = shader_cache.GetComputeKernel(code_addr);
-    const auto [program, next_bindings] = kernel->GetProgramHandle({});
+    ProgramVariant variant;
+    variant.texture_buffer_usage = SetupComputeTextures(kernel);
+    SetupComputeImages(kernel);
+
+    const auto [program, next_bindings] = kernel->GetProgramHandle(variant);
     state.draw.shader_program = program;
     state.draw.program_pipeline = 0;
 
@@ -817,13 +812,13 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     SetupComputeConstBuffers(kernel);
     SetupComputeGlobalMemory(kernel);
 
-    // TODO(Rodrigo): Bind images and samplers
-
     buffer_cache.Unmap();
 
     bind_ubo_pushbuffer.Bind();
     bind_ssbo_pushbuffer.Bind();
 
+    state.ApplyTextures();
+    state.ApplyImages();
     state.ApplyShaderProgram();
     state.ApplyProgramPipeline();
 
@@ -903,6 +898,7 @@ bool RasterizerOpenGL::AccelerateDisplay(const Tegra::FramebufferConfig& config,
     }
 
     screen_info.display_texture = surface->GetTexture();
+    screen_info.display_srgb = surface->GetSurfaceParams().srgb_conversion;
 
     return true;
 }
@@ -923,7 +919,7 @@ void RasterizerOpenGL::SetupComputeConstBuffers(const Shader& kernel) {
     const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
     for (const auto& entry : kernel->GetShaderEntries().const_buffers) {
         const auto& config = launch_desc.const_buffer_config[entry.GetIndex()];
-        const std::bitset<8> mask = launch_desc.memory_config.const_buffer_enable_mask.Value();
+        const std::bitset<8> mask = launch_desc.const_buffer_enable_mask.Value();
         Tegra::Engines::ConstBufferInfo buffer;
         buffer.address = config.Address();
         buffer.size = config.size;
@@ -982,53 +978,125 @@ void RasterizerOpenGL::SetupGlobalMemory(const GLShader::GlobalMemoryEntry& entr
     bind_ssbo_pushbuffer.Push(ssbo, buffer_offset, static_cast<GLsizeiptr>(size));
 }
 
-TextureBufferUsage RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, const Shader& shader,
-                                                   BaseBindings base_bindings) {
+TextureBufferUsage RasterizerOpenGL::SetupDrawTextures(Maxwell::ShaderStage stage,
+                                                       const Shader& shader,
+                                                       BaseBindings base_bindings) {
     MICROPROFILE_SCOPE(OpenGL_Texture);
     const auto& gpu = system.GPU();
     const auto& maxwell3d = gpu.Maxwell3D();
     const auto& entries = shader->GetShaderEntries().samplers;
 
-    ASSERT_MSG(base_bindings.sampler + entries.size() <= std::size(state.texture_units),
+    ASSERT_MSG(base_bindings.sampler + entries.size() <= std::size(state.textures),
                "Exceeded the number of active textures.");
 
     TextureBufferUsage texture_buffer_usage{0};
 
     for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
         const auto& entry = entries[bindpoint];
-        Tegra::Texture::FullTextureInfo texture;
-        if (entry.IsBindless()) {
+        const auto texture = [&]() {
+            if (!entry.IsBindless()) {
+                return maxwell3d.GetStageTexture(stage, entry.GetOffset());
+            }
             const auto cbuf = entry.GetBindlessCBuf();
             Tegra::Texture::TextureHandle tex_handle;
             tex_handle.raw = maxwell3d.AccessConstBuffer32(stage, cbuf.first, cbuf.second);
-            texture = maxwell3d.GetTextureInfo(tex_handle, entry.GetOffset());
-        } else {
-            texture = maxwell3d.GetStageTexture(stage, entry.GetOffset());
-        }
-        const u32 current_bindpoint = base_bindings.sampler + bindpoint;
+            return maxwell3d.GetTextureInfo(tex_handle, entry.GetOffset());
+        }();
 
-        auto& unit{state.texture_units[current_bindpoint]};
-        unit.sampler = sampler_cache.GetSampler(texture.tsc);
-
-        if (const auto view{texture_cache.GetTextureSurface(texture, entry)}; view) {
-            if (view->GetSurfaceParams().IsBuffer()) {
-                // Record that this texture is a texture buffer.
-                texture_buffer_usage.set(bindpoint);
-            } else {
-                // Apply swizzle to textures that are not buffers.
-                view->ApplySwizzle(texture.tic.x_source, texture.tic.y_source, texture.tic.z_source,
-                                   texture.tic.w_source);
-            }
-            state.texture_units[current_bindpoint].texture = view->GetTexture();
-        } else {
-            // Can occur when texture addr is null or its memory is unmapped/invalid
-            unit.texture = 0;
+        if (SetupTexture(base_bindings.sampler + bindpoint, texture, entry)) {
+            texture_buffer_usage.set(bindpoint);
         }
     }
 
     return texture_buffer_usage;
 }
 
+TextureBufferUsage RasterizerOpenGL::SetupComputeTextures(const Shader& kernel) {
+    MICROPROFILE_SCOPE(OpenGL_Texture);
+    const auto& compute = system.GPU().KeplerCompute();
+    const auto& entries = kernel->GetShaderEntries().samplers;
+
+    ASSERT_MSG(entries.size() <= std::size(state.textures),
+               "Exceeded the number of active textures.");
+
+    TextureBufferUsage texture_buffer_usage{0};
+
+    for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
+        const auto& entry = entries[bindpoint];
+        const auto texture = [&]() {
+            if (!entry.IsBindless()) {
+                return compute.GetTexture(entry.GetOffset());
+            }
+            const auto cbuf = entry.GetBindlessCBuf();
+            Tegra::Texture::TextureHandle tex_handle;
+            tex_handle.raw = compute.AccessConstBuffer32(cbuf.first, cbuf.second);
+            return compute.GetTextureInfo(tex_handle, entry.GetOffset());
+        }();
+
+        if (SetupTexture(bindpoint, texture, entry)) {
+            texture_buffer_usage.set(bindpoint);
+        }
+    }
+
+    return texture_buffer_usage;
+}
+
+bool RasterizerOpenGL::SetupTexture(u32 binding, const Tegra::Texture::FullTextureInfo& texture,
+                                    const GLShader::SamplerEntry& entry) {
+    state.samplers[binding] = sampler_cache.GetSampler(texture.tsc);
+
+    const auto view = texture_cache.GetTextureSurface(texture.tic, entry);
+    if (!view) {
+        // Can occur when texture addr is null or its memory is unmapped/invalid
+        state.textures[binding] = 0;
+        return false;
+    }
+    state.textures[binding] = view->GetTexture();
+
+    if (view->GetSurfaceParams().IsBuffer()) {
+        return true;
+    }
+
+    // Apply swizzle to textures that are not buffers.
+    view->ApplySwizzle(texture.tic.x_source, texture.tic.y_source, texture.tic.z_source,
+                       texture.tic.w_source);
+    return false;
+}
+
+void RasterizerOpenGL::SetupComputeImages(const Shader& shader) {
+    const auto& compute = system.GPU().KeplerCompute();
+    const auto& entries = shader->GetShaderEntries().images;
+    for (u32 bindpoint = 0; bindpoint < entries.size(); ++bindpoint) {
+        const auto& entry = entries[bindpoint];
+        const auto tic = [&]() {
+            if (!entry.IsBindless()) {
+                return compute.GetTexture(entry.GetOffset()).tic;
+            }
+            const auto cbuf = entry.GetBindlessCBuf();
+            Tegra::Texture::TextureHandle tex_handle;
+            tex_handle.raw = compute.AccessConstBuffer32(cbuf.first, cbuf.second);
+            return compute.GetTextureInfo(tex_handle, entry.GetOffset()).tic;
+        }();
+        SetupImage(bindpoint, tic, entry);
+    }
+}
+
+void RasterizerOpenGL::SetupImage(u32 binding, const Tegra::Texture::TICEntry& tic,
+                                  const GLShader::ImageEntry& entry) {
+    const auto view = texture_cache.GetImageSurface(tic, entry);
+    if (!view) {
+        state.images[binding] = 0;
+        return;
+    }
+    if (!tic.IsBuffer()) {
+        view->ApplySwizzle(tic.x_source, tic.y_source, tic.z_source, tic.w_source);
+    }
+    if (entry.IsWritten()) {
+        view->MarkAsModified(texture_cache.Tick());
+    }
+    state.images[binding] = view->GetTexture();
+}
+
 void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
     const auto& regs = system.GPU().Maxwell3D().regs;
     const bool geometry_shaders_enabled =
@@ -1119,9 +1187,12 @@ void RasterizerOpenGL::SyncStencilTestState() {
     if (!maxwell3d.dirty.stencil_test) {
         return;
     }
-    const auto& regs = maxwell3d.regs;
+    maxwell3d.dirty.stencil_test = false;
 
+    const auto& regs = maxwell3d.regs;
     state.stencil.test_enabled = regs.stencil_enable != 0;
+    state.MarkDirtyStencilState();
+
     if (!regs.stencil_enable) {
         return;
     }
@@ -1150,8 +1221,6 @@ void RasterizerOpenGL::SyncStencilTestState() {
         state.stencil.back.action_depth_fail = GL_KEEP;
         state.stencil.back.action_depth_pass = GL_KEEP;
     }
-    state.MarkDirtyStencilState();
-    maxwell3d.dirty.stencil_test = false;
 }
 
 void RasterizerOpenGL::SyncColorMask() {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -32,6 +32,7 @@
 #include "video_core/renderer_opengl/gl_state.h"
 #include "video_core/renderer_opengl/gl_texture_cache.h"
 #include "video_core/renderer_opengl/utils.h"
+#include "video_core/textures/texture.h"
 
 namespace Core {
 class System;
@@ -137,8 +138,22 @@ private:
 
     /// Configures the current textures to use for the draw command. Returns shaders texture buffer
     /// usage.
-    TextureBufferUsage SetupTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
-                                     const Shader& shader, BaseBindings base_bindings);
+    TextureBufferUsage SetupDrawTextures(Tegra::Engines::Maxwell3D::Regs::ShaderStage stage,
+                                         const Shader& shader, BaseBindings base_bindings);
+
+    /// Configures the textures used in a compute shader. Returns texture buffer usage.
+    TextureBufferUsage SetupComputeTextures(const Shader& kernel);
+
+    /// Configures a texture. Returns true when the texture is a texture buffer.
+    bool SetupTexture(u32 binding, const Tegra::Texture::FullTextureInfo& texture,
+                      const GLShader::SamplerEntry& entry);
+
+    /// Configures images in a compute shader.
+    void SetupComputeImages(const Shader& shader);
+
+    /// Configures an image.
+    void SetupImage(u32 binding, const Tegra::Texture::TICEntry& tic,
+                    const GLShader::ImageEntry& entry);
 
     /// Syncs the viewport and depth range to match the guest state
     void SyncViewport(OpenGLState& current_state);

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -348,23 +348,16 @@ Shader CachedShader::CreateKernelFromCache(const ShaderParameters& params,
 }
 
 std::tuple<GLuint, BaseBindings> CachedShader::GetProgramHandle(const ProgramVariant& variant) {
-    GLuint handle{};
-    if (program_type == ProgramType::Geometry) {
-        handle = GetGeometryShader(variant);
-    } else {
-        const auto [entry, is_cache_miss] = programs.try_emplace(variant);
-        auto& program = entry->second;
-        if (is_cache_miss) {
-            program = TryLoadProgram(variant);
-            if (!program) {
-                program = SpecializeShader(code, entries, program_type, variant);
-                disk_cache.SaveUsage(GetUsage(variant));
-            }
-
-            LabelGLObject(GL_PROGRAM, program->handle, cpu_addr);
+    const auto [entry, is_cache_miss] = programs.try_emplace(variant);
+    auto& program = entry->second;
+    if (is_cache_miss) {
+        program = TryLoadProgram(variant);
+        if (!program) {
+            program = SpecializeShader(code, entries, program_type, variant);
+            disk_cache.SaveUsage(GetUsage(variant));
         }
 
-        handle = program->handle;
+        LabelGLObject(GL_PROGRAM, program->handle, cpu_addr);
     }
 
     auto base_bindings = variant.base_bindings;
@@ -375,52 +368,9 @@ std::tuple<GLuint, BaseBindings> CachedShader::GetProgramHandle(const ProgramVar
     base_bindings.gmem += static_cast<u32>(entries.global_memory_entries.size());
     base_bindings.sampler += static_cast<u32>(entries.samplers.size());
 
-    return {handle, base_bindings};
+    return {program->handle, base_bindings};
 }
 
-GLuint CachedShader::GetGeometryShader(const ProgramVariant& variant) {
-    const auto [entry, is_cache_miss] = geometry_programs.try_emplace(variant);
-    auto& programs = entry->second;
-
-    switch (variant.primitive_mode) {
-    case GL_POINTS:
-        return LazyGeometryProgram(programs.points, variant);
-    case GL_LINES:
-    case GL_LINE_STRIP:
-        return LazyGeometryProgram(programs.lines, variant);
-    case GL_LINES_ADJACENCY:
-    case GL_LINE_STRIP_ADJACENCY:
-        return LazyGeometryProgram(programs.lines_adjacency, variant);
-    case GL_TRIANGLES:
-    case GL_TRIANGLE_STRIP:
-    case GL_TRIANGLE_FAN:
-        return LazyGeometryProgram(programs.triangles, variant);
-    case GL_TRIANGLES_ADJACENCY:
-    case GL_TRIANGLE_STRIP_ADJACENCY:
-        return LazyGeometryProgram(programs.triangles_adjacency, variant);
-    default:
-        UNREACHABLE_MSG("Unknown primitive mode.");
-        return LazyGeometryProgram(programs.points, variant);
-    }
-}
-
-GLuint CachedShader::LazyGeometryProgram(CachedProgram& target_program,
-                                         const ProgramVariant& variant) {
-    if (target_program) {
-        return target_program->handle;
-    }
-    const auto [glsl_name, debug_name, vertices] = GetPrimitiveDescription(variant.primitive_mode);
-    target_program = TryLoadProgram(variant);
-    if (!target_program) {
-        target_program = SpecializeShader(code, entries, program_type, variant);
-        disk_cache.SaveUsage(GetUsage(variant));
-    }
-
-    LabelGLObject(GL_PROGRAM, target_program->handle, cpu_addr, debug_name);
-
-    return target_program->handle;
-};
-
 CachedProgram CachedShader::TryLoadProgram(const ProgramVariant& variant) const {
     const auto found = precompiled_programs.find(GetUsage(variant));
     if (found == precompiled_programs.end()) {

src/video_core/renderer_opengl/gl_shader_cache.h

@@ -86,22 +86,6 @@ private:
     explicit CachedShader(const ShaderParameters& params, ProgramType program_type,
                           GLShader::ProgramResult result);
 
-    // Geometry programs. These are needed because GLSL needs an input topology but it's not
-    // declared by the hardware. Workaround this issue by generating a different shader per input
-    // topology class.
-    struct GeometryPrograms {
-        CachedProgram points;
-        CachedProgram lines;
-        CachedProgram lines_adjacency;
-        CachedProgram triangles;
-        CachedProgram triangles_adjacency;
-    };
-
-    GLuint GetGeometryShader(const ProgramVariant& variant);
-
-    /// Generates a geometry shader or returns one that already exists.
-    GLuint LazyGeometryProgram(CachedProgram& target_program, const ProgramVariant& variant);
-
     CachedProgram TryLoadProgram(const ProgramVariant& variant) const;
 
     ShaderDiskCacheUsage GetUsage(const ProgramVariant& variant) const;
@@ -117,11 +101,6 @@ private:
     std::size_t shader_length{};
 
     std::unordered_map<ProgramVariant, CachedProgram> programs;
-    std::unordered_map<ProgramVariant, GeometryPrograms> geometry_programs;
-
-    std::unordered_map<u32, GLuint> cbuf_resource_cache;
-    std::unordered_map<u32, GLuint> gmem_resource_cache;
-    std::unordered_map<u32, GLint> uniform_cache;
 };
 
 class ShaderCacheOpenGL final : public RasterizerCache<Shader> {

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -325,7 +325,6 @@ public:
         DeclareRegisters();
         DeclarePredicates();
         DeclareLocalMemory();
-        DeclareSharedMemory();
         DeclareInternalFlags();
         DeclareInputAttributes();
         DeclareOutputAttributes();
@@ -390,11 +389,10 @@ public:
         for (const auto& sampler : ir.GetSamplers()) {
             entries.samplers.emplace_back(sampler);
         }
-        for (const auto& image : ir.GetImages()) {
+        for (const auto& [offset, image] : ir.GetImages()) {
             entries.images.emplace_back(image);
         }
-        for (const auto& gmem_pair : ir.GetGlobalMemory()) {
-            const auto& [base, usage] = gmem_pair;
+        for (const auto& [base, usage] : ir.GetGlobalMemory()) {
             entries.global_memory_entries.emplace_back(base.cbuf_index, base.cbuf_offset,
                                                        usage.is_read, usage.is_written);
         }
@@ -501,13 +499,6 @@ private:
         code.AddNewLine();
     }
 
-    void DeclareSharedMemory() {
-        if (stage != ProgramType::Compute) {
-            return;
-        }
-        code.AddLine("shared uint {}[];", GetSharedMemory());
-    }
-
     void DeclareInternalFlags() {
         for (u32 flag = 0; flag < static_cast<u32>(InternalFlag::Amount); flag++) {
             const auto flag_code = static_cast<InternalFlag>(flag);
@@ -714,7 +705,7 @@ private:
 
     void DeclareImages() {
         const auto& images{ir.GetImages()};
-        for (const auto& image : images) {
+        for (const auto& [offset, image] : images) {
             const std::string image_type = [&]() {
                 switch (image.GetType()) {
                 case Tegra::Shader::ImageType::Texture1D:
@@ -734,9 +725,16 @@ private:
                     return "image1D";
                 }
             }();
-            code.AddLine("layout (binding = IMAGE_BINDING_{}) coherent volatile writeonly uniform "
+            std::string qualifier = "coherent volatile";
+            if (image.IsRead() && !image.IsWritten()) {
+                qualifier += " readonly";
+            } else if (image.IsWritten() && !image.IsRead()) {
+                qualifier += " writeonly";
+            }
+
+            code.AddLine("layout (binding = IMAGE_BINDING_{}) {} uniform "
                          "{} {};",
-                         image.GetIndex(), image_type, GetImage(image));
+                         image.GetIndex(), qualifier, image_type, GetImage(image));
         }
         if (!images.empty()) {
             code.AddNewLine();
@@ -866,12 +864,6 @@ private:
                 Type::Uint};
         }
 
-        if (const auto smem = std::get_if<SmemNode>(&*node)) {
-            return {
-                fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
-                Type::Uint};
-        }
-
         if (const auto internal_flag = std::get_if<InternalFlagNode>(&*node)) {
             return {GetInternalFlag(internal_flag->GetFlag()), Type::Bool};
         }
@@ -1209,11 +1201,6 @@ private:
             target = {
                 fmt::format("{}[{} >> 2]", GetLocalMemory(), Visit(lmem->GetAddress()).AsUint()),
                 Type::Uint};
-        } else if (const auto smem = std::get_if<SmemNode>(&*dest)) {
-            ASSERT(stage == ProgramType::Compute);
-            target = {
-                fmt::format("{}[{} >> 2]", GetSharedMemory(), Visit(smem->GetAddress()).AsUint()),
-                Type::Uint};
         } else if (const auto gmem = std::get_if<GmemNode>(&*dest)) {
             const std::string real = Visit(gmem->GetRealAddress()).AsUint();
             const std::string base = Visit(gmem->GetBaseAddress()).AsUint();
@@ -2095,10 +2082,6 @@ private:
         return "lmem_" + suffix;
     }
 
-    std::string GetSharedMemory() const {
-        return fmt::format("smem_{}", suffix);
-    }
-
     std::string GetInternalFlag(InternalFlag flag) const {
         constexpr std::array InternalFlagNames = {"zero_flag", "sign_flag", "carry_flag",
                                                   "overflow_flag"};

src/video_core/renderer_opengl/gl_shader_disk_cache.cpp

@@ -341,13 +341,16 @@ std::optional<ShaderDiskCacheDecompiled> ShaderDiskCacheOpenGL::LoadDecompiledEn
         u64 index{};
         u32 type{};
         u8 is_bindless{};
+        u8 is_read{};
+        u8 is_written{};
         if (!LoadObjectFromPrecompiled(offset) || !LoadObjectFromPrecompiled(index) ||
-            !LoadObjectFromPrecompiled(type) || !LoadObjectFromPrecompiled(is_bindless)) {
+            !LoadObjectFromPrecompiled(type) || !LoadObjectFromPrecompiled(is_bindless) ||
+            !LoadObjectFromPrecompiled(is_read) || !LoadObjectFromPrecompiled(is_written)) {
             return {};
         }
-        entry.entries.images.emplace_back(
-            static_cast<std::size_t>(offset), static_cast<std::size_t>(index),
-            static_cast<Tegra::Shader::ImageType>(type), is_bindless != 0);
+        entry.entries.images.emplace_back(static_cast<u64>(offset), static_cast<std::size_t>(index),
+                                          static_cast<Tegra::Shader::ImageType>(type),
+                                          is_bindless != 0, is_written != 0, is_read != 0);
     }
 
     u32 global_memory_count{};
@@ -429,7 +432,9 @@ bool ShaderDiskCacheOpenGL::SaveDecompiledFile(u64 unique_identifier, const std:
         if (!SaveObjectToPrecompiled(static_cast<u64>(image.GetOffset())) ||
             !SaveObjectToPrecompiled(static_cast<u64>(image.GetIndex())) ||
             !SaveObjectToPrecompiled(static_cast<u32>(image.GetType())) ||
-            !SaveObjectToPrecompiled(static_cast<u8>(image.IsBindless() ? 1 : 0))) {
+            !SaveObjectToPrecompiled(static_cast<u8>(image.IsBindless() ? 1 : 0)) ||
+            !SaveObjectToPrecompiled(static_cast<u8>(image.IsRead() ? 1 : 0)) ||
+            !SaveObjectToPrecompiled(static_cast<u8>(image.IsWritten() ? 1 : 0))) {
             return false;
         }
     }

src/video_core/renderer_opengl/gl_state.cpp

@@ -16,7 +16,6 @@ namespace OpenGL {
 using Maxwell = Tegra::Engines::Maxwell3D::Regs;
 
 OpenGLState OpenGLState::cur_state;
-bool OpenGLState::s_rgb_used;
 
 namespace {
 
@@ -34,6 +33,25 @@ bool UpdateTie(T1 current_value, const T2 new_value) {
     return changed;
 }
 
+template <typename T>
+std::optional<std::pair<GLuint, GLsizei>> UpdateArray(T& current_values, const T& new_values) {
+    std::optional<std::size_t> first;
+    std::size_t last;
+    for (std::size_t i = 0; i < std::size(current_values); ++i) {
+        if (!UpdateValue(current_values[i], new_values[i])) {
+            continue;
+        }
+        if (!first) {
+            first = i;
+        }
+        last = i;
+    }
+    if (!first) {
+        return std::nullopt;
+    }
+    return std::make_pair(static_cast<GLuint>(*first), static_cast<GLsizei>(last - *first + 1));
+}
+
 void Enable(GLenum cap, bool enable) {
     if (enable) {
         glEnable(cap);
@@ -134,10 +152,6 @@ OpenGLState::OpenGLState() {
     logic_op.enabled = false;
     logic_op.operation = GL_COPY;
 
-    for (auto& texture_unit : texture_units) {
-        texture_unit.Reset();
-    }
-
     draw.read_framebuffer = 0;
     draw.draw_framebuffer = 0;
     draw.vertex_array = 0;
@@ -267,8 +281,6 @@ void OpenGLState::ApplySRgb() const {
         return;
     cur_state.framebuffer_srgb.enabled = framebuffer_srgb.enabled;
     if (framebuffer_srgb.enabled) {
-        // Track if sRGB is used
-        s_rgb_used = true;
         glEnable(GL_FRAMEBUFFER_SRGB);
     } else {
         glDisable(GL_FRAMEBUFFER_SRGB);
@@ -496,52 +508,20 @@ void OpenGLState::ApplyAlphaTest() const {
 }
 
 void OpenGLState::ApplyTextures() const {
-    bool has_delta{};
-    std::size_t first{};
-    std::size_t last{};
-    std::array<GLuint, Maxwell::NumTextureSamplers> textures;
-
-    for (std::size_t i = 0; i < std::size(texture_units); ++i) {
-        const auto& texture_unit = texture_units[i];
-        auto& cur_state_texture_unit = cur_state.texture_units[i];
-        textures[i] = texture_unit.texture;
-        if (cur_state_texture_unit.texture == textures[i]) {
-            continue;
-        }
-        cur_state_texture_unit.texture = textures[i];
-        if (!has_delta) {
-            first = i;
-            has_delta = true;
-        }
-        last = i;
-    }
-    if (has_delta) {
-        glBindTextures(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
-                       textures.data() + first);
+    if (const auto update = UpdateArray(cur_state.textures, textures)) {
+        glBindTextures(update->first, update->second, textures.data() + update->first);
     }
 }
 
 void OpenGLState::ApplySamplers() const {
-    bool has_delta{};
-    std::size_t first{};
-    std::size_t last{};
-    std::array<GLuint, Maxwell::NumTextureSamplers> samplers;
-
-    for (std::size_t i = 0; i < std::size(samplers); ++i) {
-        samplers[i] = texture_units[i].sampler;
-        if (cur_state.texture_units[i].sampler == texture_units[i].sampler) {
-            continue;
-        }
-        cur_state.texture_units[i].sampler = texture_units[i].sampler;
-        if (!has_delta) {
-            first = i;
-            has_delta = true;
-        }
-        last = i;
+    if (const auto update = UpdateArray(cur_state.samplers, samplers)) {
+        glBindSamplers(update->first, update->second, samplers.data() + update->first);
     }
-    if (has_delta) {
-        glBindSamplers(static_cast<GLuint>(first), static_cast<GLsizei>(last - first + 1),
-                       samplers.data() + first);
+}
+
+void OpenGLState::ApplyImages() const {
+    if (const auto update = UpdateArray(cur_state.images, images)) {
+        glBindImageTextures(update->first, update->second, images.data() + update->first);
     }
 }
 
@@ -576,6 +556,7 @@ void OpenGLState::Apply() {
     ApplyLogicOp();
     ApplyTextures();
     ApplySamplers();
+    ApplyImages();
     if (dirty.polygon_offset) {
         ApplyPolygonOffset();
         dirty.polygon_offset = false;
@@ -606,18 +587,18 @@ void OpenGLState::EmulateViewportWithScissor() {
 }
 
 OpenGLState& OpenGLState::UnbindTexture(GLuint handle) {
-    for (auto& unit : texture_units) {
-        if (unit.texture == handle) {
-            unit.Unbind();
+    for (auto& texture : textures) {
+        if (texture == handle) {
+            texture = 0;
         }
     }
     return *this;
 }
 
 OpenGLState& OpenGLState::ResetSampler(GLuint handle) {
-    for (auto& unit : texture_units) {
-        if (unit.sampler == handle) {
-            unit.sampler = 0;
+    for (auto& sampler : samplers) {
+        if (sampler == handle) {
+            sampler = 0;
         }
     }
     return *this;

src/video_core/renderer_opengl/gl_state.h

@@ -118,21 +118,9 @@ public:
         GLenum operation;
     } logic_op;
 
-    // 3 texture units - one for each that is used in PICA fragment shader emulation
-    struct TextureUnit {
-        GLuint texture; // GL_TEXTURE_BINDING_2D
-        GLuint sampler; // GL_SAMPLER_BINDING
-
-        void Unbind() {
-            texture = 0;
-        }
-
-        void Reset() {
-            Unbind();
-            sampler = 0;
-        }
-    };
-    std::array<TextureUnit, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> texture_units;
+    std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> textures{};
+    std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::NumTextureSamplers> samplers{};
+    std::array<GLuint, Tegra::Engines::Maxwell3D::Regs::NumImages> images{};
 
     struct {
         GLuint read_framebuffer; // GL_READ_FRAMEBUFFER_BINDING
@@ -187,14 +175,6 @@ public:
         return cur_state;
     }
 
-    static bool GetsRGBUsed() {
-        return s_rgb_used;
-    }
-
-    static void ClearsRGBUsed() {
-        s_rgb_used = false;
-    }
-
     void SetDefaultViewports();
     /// Apply this state as the current OpenGL state
     void Apply();
@@ -220,6 +200,7 @@ public:
     void ApplyLogicOp() const;
     void ApplyTextures() const;
     void ApplySamplers() const;
+    void ApplyImages() const;
     void ApplyDepthClamp() const;
     void ApplyPolygonOffset() const;
     void ApplyAlphaTest() const;
@@ -264,8 +245,6 @@ public:
 private:
     static OpenGLState cur_state;
 
-    // Workaround for sRGB problems caused by QT not supporting srgb output
-    static bool s_rgb_used;
     struct {
         bool blend_state;
         bool stencil_state;

src/video_core/renderer_opengl/gl_texture_cache.h

@@ -78,6 +78,17 @@ public:
     /// Attaches this texture view to the current bound GL_DRAW_FRAMEBUFFER
     void Attach(GLenum attachment, GLenum target) const;
 
+    void ApplySwizzle(Tegra::Texture::SwizzleSource x_source,
+                      Tegra::Texture::SwizzleSource y_source,
+                      Tegra::Texture::SwizzleSource z_source,
+                      Tegra::Texture::SwizzleSource w_source);
+
+    void DecorateViewName(GPUVAddr gpu_addr, std::string prefix);
+
+    void MarkAsModified(u64 tick) {
+        surface.MarkAsModified(true, tick);
+    }
+
     GLuint GetTexture() const {
         if (is_proxy) {
             return surface.GetTexture();
@@ -89,13 +100,6 @@ public:
         return surface.GetSurfaceParams();
     }
 
-    void ApplySwizzle(Tegra::Texture::SwizzleSource x_source,
-                      Tegra::Texture::SwizzleSource y_source,
-                      Tegra::Texture::SwizzleSource z_source,
-                      Tegra::Texture::SwizzleSource w_source);
-
-    void DecorateViewName(GPUVAddr gpu_addr, std::string prefix);
-
 private:
     u32 EncodeSwizzle(Tegra::Texture::SwizzleSource x_source,
                       Tegra::Texture::SwizzleSource y_source,
@@ -111,8 +115,8 @@ private:
     GLenum target{};
 
     OGLTextureView texture_view;
-    u32 swizzle;
-    bool is_proxy;
+    u32 swizzle{};
+    bool is_proxy{};
 };
 
 class TextureCacheOpenGL final : public TextureCacheBase {

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -264,7 +264,6 @@ void RendererOpenGL::CreateRasterizer() {
     if (rasterizer) {
         return;
     }
-    OpenGLState::ClearsRGBUsed();
     rasterizer = std::make_unique<RasterizerOpenGL>(system, emu_window, screen_info);
 }
 
@@ -342,21 +341,17 @@ void RendererOpenGL::DrawScreenTriangles(const ScreenInfo& screen_info, float x,
         ScreenRectVertex(x + w, y + h, texcoords.bottom * scale_u, right * scale_v),
     }};
 
-    state.texture_units[0].texture = screen_info.display_texture;
-    // Workaround brigthness problems in SMO by enabling sRGB in the final output
-    // if it has been used in the frame. Needed because of this bug in QT: QTBUG-50987
-    state.framebuffer_srgb.enabled = OpenGLState::GetsRGBUsed();
+    state.textures[0] = screen_info.display_texture;
+    state.framebuffer_srgb.enabled = screen_info.display_srgb;
     state.AllDirty();
     state.Apply();
     glNamedBufferSubData(vertex_buffer.handle, 0, sizeof(vertices), vertices.data());
     glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
     // Restore default state
     state.framebuffer_srgb.enabled = false;
-    state.texture_units[0].texture = 0;
+    state.textures[0] = 0;
     state.AllDirty();
     state.Apply();
-    // Clear sRGB state for the next frame
-    OpenGLState::ClearsRGBUsed();
 }
 
 /**
@@ -406,8 +401,8 @@ void RendererOpenGL::CaptureScreenshot() {
     GLuint renderbuffer;
     glGenRenderbuffers(1, &renderbuffer);
     glBindRenderbuffer(GL_RENDERBUFFER, renderbuffer);
-    glRenderbufferStorage(GL_RENDERBUFFER, state.GetsRGBUsed() ? GL_SRGB8 : GL_RGB8, layout.width,
-                          layout.height);
+    glRenderbufferStorage(GL_RENDERBUFFER, screen_info.display_srgb ? GL_SRGB8 : GL_RGB8,
+                          layout.width, layout.height);
     glFramebufferRenderbuffer(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_RENDERBUFFER, renderbuffer);
 
     DrawScreen(layout);

src/video_core/renderer_opengl/renderer_opengl.h

@@ -38,7 +38,8 @@ struct TextureInfo {
 
 /// Structure used for storing information about the display target for the Switch screen
 struct ScreenInfo {
-    GLuint display_texture;
+    GLuint display_texture{};
+    bool display_srgb{};
     const Common::Rectangle<float> display_texcoords{0.0f, 0.0f, 1.0f, 1.0f};
     TextureInfo texture;
 };

src/video_core/shader/decode/image.cpp

@@ -61,56 +61,54 @@ u32 ShaderIR::DecodeImage(NodeBlock& bb, u32 pc) {
         }
 
         const auto type{instr.sust.image_type};
-        const auto& image{instr.sust.is_immediate ? GetImage(instr.image, type)
-                                                  : GetBindlessImage(instr.gpr39, type)};
+        auto& image{instr.sust.is_immediate ? GetImage(instr.image, type)
+                                            : GetBindlessImage(instr.gpr39, type)};
+        image.MarkWrite();
+
         MetaImage meta{image, values};
         const Node store{Operation(OperationCode::ImageStore, meta, std::move(coords))};
         bb.push_back(store);
         break;
     }
     default:
-        UNIMPLEMENTED_MSG("Unhandled conversion instruction: {}", opcode->get().GetName());
+        UNIMPLEMENTED_MSG("Unhandled image instruction: {}", opcode->get().GetName());
     }
 
     return pc;
 }
 
-const Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type) {
-    const auto offset{static_cast<std::size_t>(image.index.Value())};
+Image& ShaderIR::GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type) {
+    const auto offset{static_cast<u64>(image.index.Value())};
 
     // If this image has already been used, return the existing mapping.
-    const auto itr{std::find_if(used_images.begin(), used_images.end(),
-                                [=](const Image& entry) { return entry.GetOffset() == offset; })};
-    if (itr != used_images.end()) {
-        ASSERT(itr->GetType() == type);
-        return *itr;
+    const auto it = used_images.find(offset);
+    if (it != used_images.end()) {
+        ASSERT(it->second.GetType() == type);
+        return it->second;
     }
 
     // Otherwise create a new mapping for this image.
     const std::size_t next_index{used_images.size()};
-    const Image entry{offset, next_index, type};
-    return *used_images.emplace(entry).first;
+    return used_images.emplace(offset, Image{offset, next_index, type}).first->second;
 }
 
-const Image& ShaderIR::GetBindlessImage(Tegra::Shader::Register reg,
-                                        Tegra::Shader::ImageType type) {
+Image& ShaderIR::GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type) {
     const Node image_register{GetRegister(reg)};
     const auto [base_image, cbuf_index, cbuf_offset]{
         TrackCbuf(image_register, global_code, static_cast<s64>(global_code.size()))};
     const auto cbuf_key{(static_cast<u64>(cbuf_index) << 32) | static_cast<u64>(cbuf_offset)};
 
     // If this image has already been used, return the existing mapping.
-    const auto itr{std::find_if(used_images.begin(), used_images.end(),
-                                [=](const Image& entry) { return entry.GetOffset() == cbuf_key; })};
-    if (itr != used_images.end()) {
-        ASSERT(itr->GetType() == type);
-        return *itr;
+    const auto it = used_images.find(cbuf_key);
+    if (it != used_images.end()) {
+        ASSERT(it->second.GetType() == type);
+        return it->second;
     }
 
     // Otherwise create a new mapping for this image.
     const std::size_t next_index{used_images.size()};
-    const Image entry{cbuf_index, cbuf_offset, next_index, type};
-    return *used_images.emplace(entry).first;
+    return used_images.emplace(cbuf_key, Image{cbuf_index, cbuf_offset, next_index, type})
+        .first->second;
 }
 
 } // namespace VideoCommon::Shader

src/video_core/shader/decode/memory.cpp

@@ -35,7 +35,7 @@ u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) {
         return 1;
     }
 }
-} // Anonymous namespace
+} // namespace
 
 u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
@@ -106,17 +106,16 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
         }
         break;
     }
-    case OpCode::Id::LD_L:
-        LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}", static_cast<u64>(instr.ld_l.unknown));
-        [[fallthrough]];
-    case OpCode::Id::LD_S: {
-        const auto GetMemory = [&](s32 offset) {
+    case OpCode::Id::LD_L: {
+        LOG_DEBUG(HW_GPU, "LD_L cache management mode: {}",
+                  static_cast<u64>(instr.ld_l.unknown.Value()));
+
+        const auto GetLmem = [&](s32 offset) {
             ASSERT(offset % 4 == 0);
             const Node immediate_offset = Immediate(static_cast<s32>(instr.smem_imm) + offset);
             const Node address = Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8),
                                            immediate_offset);
-            return opcode->get().GetId() == OpCode::Id::LD_S ? GetSharedMemory(address)
-                                                             : GetLocalMemory(address);
+            return GetLocalMemory(address);
         };
 
         switch (instr.ldst_sl.type.Value()) {
@@ -136,16 +135,14 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
                     return 0;
                 }
             }();
-            for (u32 i = 0; i < count; ++i) {
-                SetTemporary(bb, i, GetMemory(i * 4));
-            }
-            for (u32 i = 0; i < count; ++i) {
+            for (u32 i = 0; i < count; ++i)
+                SetTemporary(bb, i, GetLmem(i * 4));
+            for (u32 i = 0; i < count; ++i)
                 SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
-            }
             break;
         }
         default:
-            UNIMPLEMENTED_MSG("{} Unhandled type: {}", opcode->get().GetName(),
+            UNIMPLEMENTED_MSG("LD_L Unhandled type: {}",
                               static_cast<u32>(instr.ldst_sl.type.Value()));
         }
         break;
@@ -212,34 +209,27 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
 
         break;
     }
-    case OpCode::Id::ST_L:
+    case OpCode::Id::ST_L: {
         LOG_DEBUG(HW_GPU, "ST_L cache management mode: {}",
                   static_cast<u64>(instr.st_l.cache_management.Value()));
-        [[fallthrough]];
-    case OpCode::Id::ST_S: {
-        const auto GetAddress = [&](s32 offset) {
+
+        const auto GetLmemAddr = [&](s32 offset) {
             ASSERT(offset % 4 == 0);
             const Node immediate = Immediate(static_cast<s32>(instr.smem_imm) + offset);
             return Operation(OperationCode::IAdd, NO_PRECISE, GetRegister(instr.gpr8), immediate);
         };
 
-        const auto set_memory = opcode->get().GetId() == OpCode::Id::ST_L
-                                    ? &ShaderIR::SetLocalMemory
-                                    : &ShaderIR::SetSharedMemory;
-
         switch (instr.ldst_sl.type.Value()) {
         case Tegra::Shader::StoreType::Bits128:
-            (this->*set_memory)(bb, GetAddress(12), GetRegister(instr.gpr0.Value() + 3));
-            (this->*set_memory)(bb, GetAddress(8), GetRegister(instr.gpr0.Value() + 2));
-            [[fallthrough]];
+            SetLocalMemory(bb, GetLmemAddr(12), GetRegister(instr.gpr0.Value() + 3));
+            SetLocalMemory(bb, GetLmemAddr(8), GetRegister(instr.gpr0.Value() + 2));
         case Tegra::Shader::StoreType::Bits64:
-            (this->*set_memory)(bb, GetAddress(4), GetRegister(instr.gpr0.Value() + 1));
-            [[fallthrough]];
+            SetLocalMemory(bb, GetLmemAddr(4), GetRegister(instr.gpr0.Value() + 1));
         case Tegra::Shader::StoreType::Bits32:
-            (this->*set_memory)(bb, GetAddress(0), GetRegister(instr.gpr0));
+            SetLocalMemory(bb, GetLmemAddr(0), GetRegister(instr.gpr0));
             break;
         default:
-            UNIMPLEMENTED_MSG("{} unhandled type: {}", opcode->get().GetName(),
+            UNIMPLEMENTED_MSG("ST_L Unhandled type: {}",
                               static_cast<u32>(instr.ldst_sl.type.Value()));
         }
         break;

src/video_core/shader/decode/shift.cpp

@@ -17,8 +17,8 @@ u32 ShaderIR::DecodeShift(NodeBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
-    const Node op_a = GetRegister(instr.gpr8);
-    const Node op_b = [&]() {
+    Node op_a = GetRegister(instr.gpr8);
+    Node op_b = [&]() {
         if (instr.is_b_imm) {
             return Immediate(instr.alu.GetSignedImm20_20());
         } else if (instr.is_b_gpr) {
@@ -32,16 +32,23 @@ u32 ShaderIR::DecodeShift(NodeBlock& bb, u32 pc) {
     case OpCode::Id::SHR_C:
     case OpCode::Id::SHR_R:
     case OpCode::Id::SHR_IMM: {
-        const Node value = SignedOperation(OperationCode::IArithmeticShiftRight,
-                                           instr.shift.is_signed, PRECISE, op_a, op_b);
+        if (instr.shr.wrap) {
+            op_b = Operation(OperationCode::UBitwiseAnd, std::move(op_b), Immediate(0x1f));
+        } else {
+            op_b = Operation(OperationCode::IMax, std::move(op_b), Immediate(0));
+            op_b = Operation(OperationCode::IMin, std::move(op_b), Immediate(31));
+        }
+
+        Node value = SignedOperation(OperationCode::IArithmeticShiftRight, instr.shift.is_signed,
+                                     std::move(op_a), std::move(op_b));
         SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
-        SetRegister(bb, instr.gpr0, value);
+        SetRegister(bb, instr.gpr0, std::move(value));
         break;
     }
     case OpCode::Id::SHL_C:
     case OpCode::Id::SHL_R:
     case OpCode::Id::SHL_IMM: {
-        const Node value = Operation(OperationCode::ILogicalShiftLeft, PRECISE, op_a, op_b);
+        const Node value = Operation(OperationCode::ILogicalShiftLeft, op_a, op_b);
         SetInternalFlagsFromInteger(bb, value, instr.generates_cc);
         SetRegister(bb, instr.gpr0, value);
         break;

src/video_core/shader/node.h

@@ -198,13 +198,12 @@ class PredicateNode;
 class AbufNode;
 class CbufNode;
 class LmemNode;
-class SmemNode;
 class GmemNode;
 class CommentNode;
 
 using NodeData =
     std::variant<OperationNode, ConditionalNode, GprNode, ImmediateNode, InternalFlagNode,
-                 PredicateNode, AbufNode, CbufNode, LmemNode, SmemNode, GmemNode, CommentNode>;
+                 PredicateNode, AbufNode, CbufNode, LmemNode, GmemNode, CommentNode>;
 using Node = std::shared_ptr<NodeData>;
 using Node4 = std::array<Node, 4>;
 using NodeBlock = std::vector<Node>;
@@ -274,46 +273,64 @@ private:
     bool is_bindless{}; ///< Whether this sampler belongs to a bindless texture or not.
 };
 
-class Image {
+class Image final {
 public:
-    explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type)
+    constexpr explicit Image(u64 offset, std::size_t index, Tegra::Shader::ImageType type)
         : offset{offset}, index{index}, type{type}, is_bindless{false} {}
 
-    explicit Image(u32 cbuf_index, u32 cbuf_offset, std::size_t index,
-                   Tegra::Shader::ImageType type)
+    constexpr explicit Image(u32 cbuf_index, u32 cbuf_offset, std::size_t index,
+                             Tegra::Shader::ImageType type)
         : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type},
           is_bindless{true} {}
 
-    explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type,
-                   bool is_bindless)
-        : offset{offset}, index{index}, type{type}, is_bindless{is_bindless} {}
+    constexpr explicit Image(std::size_t offset, std::size_t index, Tegra::Shader::ImageType type,
+                             bool is_bindless, bool is_written, bool is_read)
+        : offset{offset}, index{index}, type{type}, is_bindless{is_bindless},
+          is_written{is_written}, is_read{is_read} {}
 
-    std::size_t GetOffset() const {
+    void MarkRead() {
+        is_read = true;
+    }
+
+    void MarkWrite() {
+        is_written = true;
+    }
+
+    constexpr std::size_t GetOffset() const {
         return offset;
     }
 
-    std::size_t GetIndex() const {
+    constexpr std::size_t GetIndex() const {
         return index;
     }
 
-    Tegra::Shader::ImageType GetType() const {
+    constexpr Tegra::Shader::ImageType GetType() const {
         return type;
     }
 
-    bool IsBindless() const {
+    constexpr bool IsBindless() const {
         return is_bindless;
     }
 
-    bool operator<(const Image& rhs) const {
-        return std::tie(offset, index, type, is_bindless) <
-               std::tie(rhs.offset, rhs.index, rhs.type, rhs.is_bindless);
+    constexpr bool IsRead() const {
+        return is_read;
+    }
+
+    constexpr bool IsWritten() const {
+        return is_written;
+    }
+
+    constexpr std::pair<u32, u32> GetBindlessCBuf() const {
+        return {static_cast<u32>(offset >> 32), static_cast<u32>(offset)};
     }
 
 private:
-    std::size_t offset{};
+    u64 offset{};
     std::size_t index{};
     Tegra::Shader::ImageType type{};
     bool is_bindless{};
+    bool is_read{};
+    bool is_written{};
 };
 
 struct GlobalMemoryBase {
@@ -537,19 +554,6 @@ private:
     Node address;
 };
 
-/// Shared memory node
-class SmemNode final {
-public:
-    explicit SmemNode(Node address) : address{std::move(address)} {}
-
-    const Node& GetAddress() const {
-        return address;
-    }
-
-private:
-    Node address;
-};
-
 /// Global memory node
 class GmemNode final {
 public:

src/video_core/shader/shader_ir.cpp

@@ -137,10 +137,6 @@ Node ShaderIR::GetLocalMemory(Node address) {
     return MakeNode<LmemNode>(std::move(address));
 }
 
-Node ShaderIR::GetSharedMemory(Node address) {
-    return MakeNode<SmemNode>(std::move(address));
-}
-
 Node ShaderIR::GetTemporary(u32 id) {
     return GetRegister(Register::ZeroIndex + 1 + id);
 }
@@ -382,11 +378,6 @@ void ShaderIR::SetLocalMemory(NodeBlock& bb, Node address, Node value) {
         Operation(OperationCode::Assign, GetLocalMemory(std::move(address)), std::move(value)));
 }
 
-void ShaderIR::SetSharedMemory(NodeBlock& bb, Node address, Node value) {
-    bb.push_back(
-        Operation(OperationCode::Assign, GetSharedMemory(std::move(address)), std::move(value)));
-}
-
 void ShaderIR::SetTemporary(NodeBlock& bb, u32 id, Node value) {
     SetRegister(bb, Register::ZeroIndex + 1 + id, std::move(value));
 }

src/video_core/shader/shader_ir.h

@@ -95,7 +95,7 @@ public:
         return used_samplers;
     }
 
-    const std::set<Image>& GetImages() const {
+    const std::map<u64, Image>& GetImages() const {
         return used_images;
     }
 
@@ -208,8 +208,6 @@ private:
     Node GetInternalFlag(InternalFlag flag, bool negated = false);
     /// Generates a node representing a local memory address
     Node GetLocalMemory(Node address);
-    /// Generates a node representing a shared memory address
-    Node GetSharedMemory(Node address);
     /// Generates a temporary, internally it uses a post-RZ register
     Node GetTemporary(u32 id);
 
@@ -219,10 +217,8 @@ private:
     void SetPredicate(NodeBlock& bb, u64 dest, Node src);
     /// Sets an internal flag. src value must be a bool-evaluated node
     void SetInternalFlag(NodeBlock& bb, InternalFlag flag, Node value);
-    /// Sets a local memory address with a value.
+    /// Sets a local memory address. address and value must be a number-evaluated node
     void SetLocalMemory(NodeBlock& bb, Node address, Node value);
-    /// Sets a shared memory address with a value.
-    void SetSharedMemory(NodeBlock& bb, Node address, Node value);
     /// Sets a temporary. Internally it uses a post-RZ register
     void SetTemporary(NodeBlock& bb, u32 id, Node value);
 
@@ -276,10 +272,10 @@ private:
                                       bool is_shadow);
 
     /// Accesses an image.
-    const Image& GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type);
+    Image& GetImage(Tegra::Shader::Image image, Tegra::Shader::ImageType type);
 
     /// Access a bindless image sampler.
-    const Image& GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type);
+    Image& GetBindlessImage(Tegra::Shader::Register reg, Tegra::Shader::ImageType type);
 
     /// Extracts a sequence of bits from a node
     Node BitfieldExtract(Node value, u32 offset, u32 bits);
@@ -360,7 +356,7 @@ private:
     std::set<Tegra::Shader::Attribute::Index> used_output_attributes;
     std::map<u32, ConstBuffer> used_cbufs;
     std::set<Sampler> used_samplers;
-    std::set<Image> used_images;
+    std::map<u64, Image> used_images;
     std::array<bool, Tegra::Engines::Maxwell3D::Regs::NumClipDistances> used_clip_distances{};
     std::map<GlobalMemoryBase, GlobalMemoryUsage> used_global_memory;
     bool uses_layer{};

src/video_core/texture_cache/surface_base.h

@@ -195,18 +195,18 @@ public:
 
     virtual void DownloadTexture(std::vector<u8>& staging_buffer) = 0;
 
-    void MarkAsModified(const bool is_modified_, const u64 tick) {
+    void MarkAsModified(bool is_modified_, u64 tick) {
         is_modified = is_modified_ || is_target;
         modification_tick = tick;
     }
 
-    void MarkAsRenderTarget(const bool is_target, const u32 index) {
-        this->is_target = is_target;
-        this->index = index;
+    void MarkAsRenderTarget(bool is_target_, u32 index_) {
+        is_target = is_target_;
+        index = index_;
     }
 
-    void MarkAsPicked(const bool is_picked) {
-        this->is_picked = is_picked;
+    void MarkAsPicked(bool is_picked_) {
+        is_picked = is_picked_;
     }
 
     bool IsModified() const {

src/video_core/texture_cache/surface_params.cpp

@@ -24,55 +24,62 @@ using VideoCore::Surface::SurfaceTarget;
 using VideoCore::Surface::SurfaceTargetFromTextureType;
 using VideoCore::Surface::SurfaceType;
 
-SurfaceTarget TextureType2SurfaceTarget(Tegra::Shader::TextureType type, bool is_array) {
+namespace {
+
+SurfaceTarget TextureTypeToSurfaceTarget(Tegra::Shader::TextureType type, bool is_array) {
     switch (type) {
-    case Tegra::Shader::TextureType::Texture1D: {
-        if (is_array)
-            return SurfaceTarget::Texture1DArray;
-        else
-            return SurfaceTarget::Texture1D;
-    }
-    case Tegra::Shader::TextureType::Texture2D: {
-        if (is_array)
-            return SurfaceTarget::Texture2DArray;
-        else
-            return SurfaceTarget::Texture2D;
-    }
-    case Tegra::Shader::TextureType::Texture3D: {
+    case Tegra::Shader::TextureType::Texture1D:
+        return is_array ? SurfaceTarget::Texture1DArray : SurfaceTarget::Texture1D;
+    case Tegra::Shader::TextureType::Texture2D:
+        return is_array ? SurfaceTarget::Texture2DArray : SurfaceTarget::Texture2D;
+    case Tegra::Shader::TextureType::Texture3D:
         ASSERT(!is_array);
         return SurfaceTarget::Texture3D;
-    }
-    case Tegra::Shader::TextureType::TextureCube: {
-        if (is_array)
-            return SurfaceTarget::TextureCubeArray;
-        else
-            return SurfaceTarget::TextureCubemap;
-    }
-    default: {
+    case Tegra::Shader::TextureType::TextureCube:
+        return is_array ? SurfaceTarget::TextureCubeArray : SurfaceTarget::TextureCubemap;
+    default:
         UNREACHABLE();
         return SurfaceTarget::Texture2D;
     }
+}
+
+SurfaceTarget ImageTypeToSurfaceTarget(Tegra::Shader::ImageType type) {
+    switch (type) {
+    case Tegra::Shader::ImageType::Texture1D:
+        return SurfaceTarget::Texture1D;
+    case Tegra::Shader::ImageType::TextureBuffer:
+        return SurfaceTarget::TextureBuffer;
+    case Tegra::Shader::ImageType::Texture1DArray:
+        return SurfaceTarget::Texture1DArray;
+    case Tegra::Shader::ImageType::Texture2D:
+        return SurfaceTarget::Texture2D;
+    case Tegra::Shader::ImageType::Texture2DArray:
+        return SurfaceTarget::Texture2DArray;
+    case Tegra::Shader::ImageType::Texture3D:
+        return SurfaceTarget::Texture3D;
+    default:
+        UNREACHABLE();
+        return SurfaceTarget::Texture2D;
     }
 }
 
-namespace {
 constexpr u32 GetMipmapSize(bool uncompressed, u32 mip_size, u32 tile) {
     return uncompressed ? mip_size : std::max(1U, (mip_size + tile - 1) / tile);
 }
+
 } // Anonymous namespace
 
-SurfaceParams SurfaceParams::CreateForTexture(Core::System& system,
-                                              const Tegra::Texture::FullTextureInfo& config,
+SurfaceParams SurfaceParams::CreateForTexture(const Tegra::Texture::TICEntry& tic,
                                               const VideoCommon::Shader::Sampler& entry) {
     SurfaceParams params;
-    params.is_tiled = config.tic.IsTiled();
-    params.srgb_conversion = config.tic.IsSrgbConversionEnabled();
-    params.block_width = params.is_tiled ? config.tic.BlockWidth() : 0,
-    params.block_height = params.is_tiled ? config.tic.BlockHeight() : 0,
-    params.block_depth = params.is_tiled ? config.tic.BlockDepth() : 0,
-    params.tile_width_spacing = params.is_tiled ? (1 << config.tic.tile_width_spacing.Value()) : 1;
-    params.pixel_format = PixelFormatFromTextureFormat(config.tic.format, config.tic.r_type.Value(),
-                                                       params.srgb_conversion);
+    params.is_tiled = tic.IsTiled();
+    params.srgb_conversion = tic.IsSrgbConversionEnabled();
+    params.block_width = params.is_tiled ? tic.BlockWidth() : 0,
+    params.block_height = params.is_tiled ? tic.BlockHeight() : 0,
+    params.block_depth = params.is_tiled ? tic.BlockDepth() : 0,
+    params.tile_width_spacing = params.is_tiled ? (1 << tic.tile_width_spacing.Value()) : 1;
+    params.pixel_format =
+        PixelFormatFromTextureFormat(tic.format, tic.r_type.Value(), params.srgb_conversion);
     params.type = GetFormatType(params.pixel_format);
     if (entry.IsShadow() && params.type == SurfaceType::ColorTexture) {
         switch (params.pixel_format) {
@@ -92,31 +99,72 @@ SurfaceParams SurfaceParams::CreateForTexture(Core::System& system,
         }
         params.type = GetFormatType(params.pixel_format);
     }
-    params.component_type = ComponentTypeFromTexture(config.tic.r_type.Value());
+    params.component_type = ComponentTypeFromTexture(tic.r_type.Value());
     params.type = GetFormatType(params.pixel_format);
     // TODO: on 1DBuffer we should use the tic info.
-    if (!config.tic.IsBuffer()) {
-        params.target = TextureType2SurfaceTarget(entry.GetType(), entry.IsArray());
-        params.width = config.tic.Width();
-        params.height = config.tic.Height();
-        params.depth = config.tic.Depth();
-        params.pitch = params.is_tiled ? 0 : config.tic.Pitch();
-        if (params.target == SurfaceTarget::TextureCubemap ||
-            params.target == SurfaceTarget::TextureCubeArray) {
-            params.depth *= 6;
-        }
-        params.num_levels = config.tic.max_mip_level + 1;
-        params.emulated_levels = std::min(params.num_levels, params.MaxPossibleMipmap());
-        params.is_layered = params.IsLayered();
-    } else {
+    if (tic.IsBuffer()) {
         params.target = SurfaceTarget::TextureBuffer;
-        params.width = config.tic.Width();
+        params.width = tic.Width();
         params.pitch = params.width * params.GetBytesPerPixel();
         params.height = 1;
         params.depth = 1;
         params.num_levels = 1;
         params.emulated_levels = 1;
         params.is_layered = false;
+    } else {
+        params.target = TextureTypeToSurfaceTarget(entry.GetType(), entry.IsArray());
+        params.width = tic.Width();
+        params.height = tic.Height();
+        params.depth = tic.Depth();
+        params.pitch = params.is_tiled ? 0 : tic.Pitch();
+        if (params.target == SurfaceTarget::TextureCubemap ||
+            params.target == SurfaceTarget::TextureCubeArray) {
+            params.depth *= 6;
+        }
+        params.num_levels = tic.max_mip_level + 1;
+        params.emulated_levels = std::min(params.num_levels, params.MaxPossibleMipmap());
+        params.is_layered = params.IsLayered();
+    }
+    return params;
+}
+
+SurfaceParams SurfaceParams::CreateForImage(const Tegra::Texture::TICEntry& tic,
+                                            const VideoCommon::Shader::Image& entry) {
+    SurfaceParams params;
+    params.is_tiled = tic.IsTiled();
+    params.srgb_conversion = tic.IsSrgbConversionEnabled();
+    params.block_width = params.is_tiled ? tic.BlockWidth() : 0,
+    params.block_height = params.is_tiled ? tic.BlockHeight() : 0,
+    params.block_depth = params.is_tiled ? tic.BlockDepth() : 0,
+    params.tile_width_spacing = params.is_tiled ? (1 << tic.tile_width_spacing.Value()) : 1;
+    params.pixel_format =
+        PixelFormatFromTextureFormat(tic.format, tic.r_type.Value(), params.srgb_conversion);
+    params.type = GetFormatType(params.pixel_format);
+    params.component_type = ComponentTypeFromTexture(tic.r_type.Value());
+    params.type = GetFormatType(params.pixel_format);
+    params.target = ImageTypeToSurfaceTarget(entry.GetType());
+    // TODO: on 1DBuffer we should use the tic info.
+    if (tic.IsBuffer()) {
+        params.target = SurfaceTarget::TextureBuffer;
+        params.width = tic.Width();
+        params.pitch = params.width * params.GetBytesPerPixel();
+        params.height = 1;
+        params.depth = 1;
+        params.num_levels = 1;
+        params.emulated_levels = 1;
+        params.is_layered = false;
+    } else {
+        params.width = tic.Width();
+        params.height = tic.Height();
+        params.depth = tic.Depth();
+        params.pitch = params.is_tiled ? 0 : tic.Pitch();
+        if (params.target == SurfaceTarget::TextureCubemap ||
+            params.target == SurfaceTarget::TextureCubeArray) {
+            params.depth *= 6;
+        }
+        params.num_levels = tic.max_mip_level + 1;
+        params.emulated_levels = std::min(params.num_levels, params.MaxPossibleMipmap());
+        params.is_layered = params.IsLayered();
     }
     return params;
 }

src/video_core/texture_cache/surface_params.h

@@ -4,8 +4,6 @@
 
 #pragma once
 
-#include <map>
-
 #include "common/alignment.h"
 #include "common/bit_util.h"
 #include "common/cityhash.h"
@@ -23,10 +21,13 @@ using VideoCore::Surface::SurfaceCompression;
 class SurfaceParams {
 public:
     /// Creates SurfaceCachedParams from a texture configuration.
-    static SurfaceParams CreateForTexture(Core::System& system,
-                                          const Tegra::Texture::FullTextureInfo& config,
+    static SurfaceParams CreateForTexture(const Tegra::Texture::TICEntry& tic,
                                           const VideoCommon::Shader::Sampler& entry);
 
+    /// Creates SurfaceCachedParams from an image configuration.
+    static SurfaceParams CreateForImage(const Tegra::Texture::TICEntry& tic,
+                                        const VideoCommon::Shader::Image& entry);
+
     /// Creates SurfaceCachedParams for a depth buffer configuration.
     static SurfaceParams CreateForDepthBuffer(
         Core::System& system, u32 zeta_width, u32 zeta_height, Tegra::DepthFormat format,

src/video_core/texture_cache/surface_view.cpp

@@ -10,7 +10,7 @@
 namespace VideoCommon {
 
 std::size_t ViewParams::Hash() const {
-    return static_cast<std::size_t>(base_layer) ^ static_cast<std::size_t>(num_layers << 16) ^
+    return static_cast<std::size_t>(base_layer) ^ (static_cast<std::size_t>(num_layers) << 16) ^
            (static_cast<std::size_t>(base_level) << 24) ^
            (static_cast<std::size_t>(num_levels) << 32) ^ (static_cast<std::size_t>(target) << 36);
 }

src/video_core/texture_cache/surface_view.h

@@ -13,8 +13,8 @@
 namespace VideoCommon {
 
 struct ViewParams {
-    ViewParams(VideoCore::Surface::SurfaceTarget target, u32 base_layer, u32 num_layers,
-               u32 base_level, u32 num_levels)
+    constexpr explicit ViewParams(VideoCore::Surface::SurfaceTarget target, u32 base_layer,
+                                  u32 num_layers, u32 base_level, u32 num_levels)
         : target{target}, base_layer{base_layer}, num_layers{num_layers}, base_level{base_level},
           num_levels{num_levels} {}
 
@@ -22,12 +22,6 @@ struct ViewParams {
 
     bool operator==(const ViewParams& rhs) const;
 
-    VideoCore::Surface::SurfaceTarget target{};
-    u32 base_layer{};
-    u32 num_layers{};
-    u32 base_level{};
-    u32 num_levels{};
-
     bool IsLayered() const {
         switch (target) {
         case VideoCore::Surface::SurfaceTarget::Texture1DArray:
@@ -39,13 +33,19 @@ struct ViewParams {
             return false;
         }
     }
+
+    VideoCore::Surface::SurfaceTarget target{};
+    u32 base_layer{};
+    u32 num_layers{};
+    u32 base_level{};
+    u32 num_levels{};
 };
 
 class ViewBase {
 public:
-    ViewBase(const ViewParams& params) : params{params} {}
+    constexpr explicit ViewBase(const ViewParams& params) : params{params} {}
 
-    const ViewParams& GetViewParams() const {
+    constexpr const ViewParams& GetViewParams() const {
         return params;
     }
 

src/video_core/texture_cache/texture_cache.h

@@ -89,14 +89,29 @@ public:
         }
     }
 
-    TView GetTextureSurface(const Tegra::Texture::FullTextureInfo& config,
+    TView GetTextureSurface(const Tegra::Texture::TICEntry& tic,
                             const VideoCommon::Shader::Sampler& entry) {
         std::lock_guard lock{mutex};
-        const auto gpu_addr{config.tic.Address()};
+        const auto gpu_addr{tic.Address()};
         if (!gpu_addr) {
             return {};
         }
-        const auto params{SurfaceParams::CreateForTexture(system, config, entry)};
+        const auto params{SurfaceParams::CreateForTexture(tic, entry)};
+        const auto [surface, view] = GetSurface(gpu_addr, params, true, false);
+        if (guard_samplers) {
+            sampled_textures.push_back(surface);
+        }
+        return view;
+    }
+
+    TView GetImageSurface(const Tegra::Texture::TICEntry& tic,
+                          const VideoCommon::Shader::Image& entry) {
+        std::lock_guard lock{mutex};
+        const auto gpu_addr{tic.Address()};
+        if (!gpu_addr) {
+            return {};
+        }
+        const auto params{SurfaceParams::CreateForImage(tic, entry)};
         const auto [surface, view] = GetSurface(gpu_addr, params, true, false);
         if (guard_samplers) {
             sampled_textures.push_back(surface);