Implemented Fast Layered Copy

textures/decoders: Replace magic numbers

.travis/clang-format/build.sh

@@ -1,3 +1,3 @@
 #!/bin/bash -ex
 
-docker run -v $(pwd):/yuzu ubuntu:18.04 /bin/bash -ex /yuzu/.travis/clang-format/docker.sh
+docker run --env-file .travis/common/travis-ci.env -v $(pwd):/yuzu -v "$HOME/.ccache":/root/.ccache citraemu/build-environments:linux-clang-format /bin/bash -ex /yuzu/.travis/clang-format/docker.sh

.travis/clang-format/deps.sh

@@ -1,3 +1,3 @@
 #!/bin/sh -ex
 
-docker pull ubuntu:18.04
+docker pull citraemu/build-environments:linux-clang-format

.travis/clang-format/docker.sh

@@ -1,8 +1,5 @@
 #!/bin/bash -ex
 
-apt-get update
-apt-get install -y clang-format-6.0
-
 # Run clang-format
 cd /yuzu
 ./.travis/clang-format/script.sh

externals/cmake-modules/GetGitRevisionDescription.cmake.in

@@ -33,6 +33,10 @@ else()
 endif()
 
 if(NOT HEAD_HASH)
-	file(READ "@GIT_DATA@/head-ref" HEAD_HASH LIMIT 1024)
-	string(STRIP "${HEAD_HASH}" HEAD_HASH)
+	if(EXISTS "@GIT_DATA@/head-ref")
+		file(READ "@GIT_DATA@/head-ref" HEAD_HASH LIMIT 1024)
+		string(STRIP "${HEAD_HASH}" HEAD_HASH)
+	else()
+		set(HEAD_HASH "Unknown")
+	endif()
 endif()

src/core/hle/kernel/process.cpp

@@ -252,8 +252,8 @@ ResultCode Process::HeapFree(VAddr target, u32 size) {
     return vm_manager.HeapFree(target, size);
 }
 
-ResultCode Process::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size) {
-    return vm_manager.MirrorMemory(dst_addr, src_addr, size);
+ResultCode Process::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size, MemoryState state) {
+    return vm_manager.MirrorMemory(dst_addr, src_addr, size, state);
 }
 
 ResultCode Process::UnmapMemory(VAddr dst_addr, VAddr /*src_addr*/, u64 size) {

src/core/hle/kernel/process.h

@@ -259,7 +259,8 @@ public:
     ResultVal<VAddr> HeapAllocate(VAddr target, u64 size, VMAPermission perms);
     ResultCode HeapFree(VAddr target, u32 size);
 
-    ResultCode MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size);
+    ResultCode MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size,
+                            MemoryState state = MemoryState::Mapped);
 
     ResultCode UnmapMemory(VAddr dst_addr, VAddr src_addr, u64 size);
 

src/core/hle/kernel/svc.cpp

@@ -1181,7 +1181,7 @@ static ResultCode CloseHandle(Handle handle) {
 
 /// Reset an event
 static ResultCode ResetSignal(Handle handle) {
-    LOG_WARNING(Kernel_SVC, "(STUBBED) called handle 0x{:08X}", handle);
+    LOG_DEBUG(Kernel_SVC, "called handle 0x{:08X}", handle);
 
     const auto& handle_table = Core::CurrentProcess()->GetHandleTable();
     auto event = handle_table.Get<Event>(handle);

src/core/hle/kernel/vm_manager.cpp

@@ -298,7 +298,7 @@ ResultCode VMManager::HeapFree(VAddr target, u64 size) {
     return RESULT_SUCCESS;
 }
 
-ResultCode VMManager::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size) {
+ResultCode VMManager::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size, MemoryState state) {
     const auto vma = FindVMA(src_addr);
 
     ASSERT_MSG(vma != vma_map.end(), "Invalid memory address");
@@ -312,8 +312,8 @@ ResultCode VMManager::MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size) {
     const std::shared_ptr<std::vector<u8>>& backing_block = vma->second.backing_block;
     const std::size_t backing_block_offset = vma->second.offset + vma_offset;
 
-    CASCADE_RESULT(auto new_vma, MapMemoryBlock(dst_addr, backing_block, backing_block_offset, size,
-                                                MemoryState::Mapped));
+    CASCADE_RESULT(auto new_vma,
+                   MapMemoryBlock(dst_addr, backing_block, backing_block_offset, size, state));
     // Protect mirror with permissions from old region
     Reprotect(new_vma, vma->second.permissions);
     // Remove permissions from old region

src/core/hle/kernel/vm_manager.h

@@ -189,7 +189,8 @@ public:
     ResultVal<VAddr> HeapAllocate(VAddr target, u64 size, VMAPermission perms);
     ResultCode HeapFree(VAddr target, u64 size);
 
-    ResultCode MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size);
+    ResultCode MirrorMemory(VAddr dst_addr, VAddr src_addr, u64 size,
+                            MemoryState state = MemoryState::Mapped);
 
     /**
      * Scans all VMAs and updates the page table range of any that use the given vector as backing

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

@@ -203,8 +203,8 @@ ISelfController::ISelfController(std::shared_ptr<NVFlinger::NVFlinger> nvflinger
 ISelfController::~ISelfController() = default;
 
 void ISelfController::SetFocusHandlingMode(Kernel::HLERequestContext& ctx) {
-    // Takes 3 input u8s with each field located immediately after the previous u8, these are
-    // bool flags. No output.
+    // Takes 3 input u8s with each field located immediately after the previous
+    // u8, these are bool flags. No output.
 
     IPC::RequestParser rp{ctx};
 
@@ -258,8 +258,8 @@ void ISelfController::SetOperationModeChangedNotification(Kernel::HLERequestCont
 }
 
 void ISelfController::SetOutOfFocusSuspendingEnabled(Kernel::HLERequestContext& ctx) {
-    // Takes 3 input u8s with each field located immediately after the previous u8, these are
-    // bool flags. No output.
+    // Takes 3 input u8s with each field located immediately after the previous
+    // u8, these are bool flags. No output.
     IPC::RequestParser rp{ctx};
 
     bool enabled = rp.Pop<bool>();
@@ -302,8 +302,8 @@ void ISelfController::SetScreenShotImageOrientation(Kernel::HLERequestContext& c
 }
 
 void ISelfController::CreateManagedDisplayLayer(Kernel::HLERequestContext& ctx) {
-    // TODO(Subv): Find out how AM determines the display to use, for now just create the layer
-    // in the Default display.
+    // TODO(Subv): Find out how AM determines the display to use, for now just
+    // create the layer in the Default display.
     u64 display_id = nvflinger->OpenDisplay("Default");
     u64 layer_id = nvflinger->CreateLayer(display_id);
 
@@ -733,7 +733,7 @@ IApplicationFunctions::IApplicationFunctions() : ServiceFramework("IApplicationF
         {70, nullptr, "RequestToShutdown"},
         {71, nullptr, "RequestToReboot"},
         {80, nullptr, "ExitAndRequestToShowThanksMessage"},
-        {90, nullptr, "EnableApplicationCrashReport"},
+        {90, &IApplicationFunctions::EnableApplicationCrashReport, "EnableApplicationCrashReport"},
         {100, nullptr, "InitializeApplicationCopyrightFrameBuffer"},
         {101, nullptr, "SetApplicationCopyrightImage"},
         {102, nullptr, "SetApplicationCopyrightVisibility"},
@@ -752,6 +752,12 @@ IApplicationFunctions::IApplicationFunctions() : ServiceFramework("IApplicationF
 
 IApplicationFunctions::~IApplicationFunctions() = default;
 
+void IApplicationFunctions::EnableApplicationCrashReport(Kernel::HLERequestContext& ctx) {
+    IPC::ResponseBuilder rb{ctx, 2};
+    rb.Push(RESULT_SUCCESS);
+    LOG_WARNING(Service_AM, "(STUBBED) called");
+}
+
 void IApplicationFunctions::BeginBlockingHomeButtonShortAndLongPressed(
     Kernel::HLERequestContext& ctx) {
     IPC::ResponseBuilder rb{ctx, 2};
@@ -821,7 +827,8 @@ void IApplicationFunctions::EnsureSaveData(Kernel::HLERequestContext& ctx) {
 
 void IApplicationFunctions::SetTerminateResult(Kernel::HLERequestContext& ctx) {
     // Takes an input u32 Result, no output.
-    // For example, in some cases official apps use this with error 0x2A2 then uses svcBreak.
+    // For example, in some cases official apps use this with error 0x2A2 then
+    // uses svcBreak.
 
     IPC::RequestParser rp{ctx};
     u32 result = rp.Pop<u32>();
@@ -884,8 +891,8 @@ void IApplicationFunctions::GetPseudoDeviceId(Kernel::HLERequestContext& ctx) {
 void InstallInterfaces(SM::ServiceManager& service_manager,
                        std::shared_ptr<NVFlinger::NVFlinger> nvflinger) {
     auto message_queue = std::make_shared<AppletMessageQueue>();
-    message_queue->PushMessage(
-        AppletMessageQueue::AppletMessage::FocusStateChanged); // Needed on game boot
+    message_queue->PushMessage(AppletMessageQueue::AppletMessage::FocusStateChanged); // Needed on
+                                                                                      // game boot
 
     std::make_shared<AppletAE>(nvflinger, message_queue)->InstallAsService(service_manager);
     std::make_shared<AppletOE>(nvflinger, message_queue)->InstallAsService(service_manager);

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

@@ -185,6 +185,7 @@ private:
     void EndBlockingHomeButtonShortAndLongPressed(Kernel::HLERequestContext& ctx);
     void BeginBlockingHomeButton(Kernel::HLERequestContext& ctx);
     void EndBlockingHomeButton(Kernel::HLERequestContext& ctx);
+    void EnableApplicationCrashReport(Kernel::HLERequestContext& ctx);
 };
 
 class IHomeMenuFunctions final : public ServiceFramework<IHomeMenuFunctions> {

src/core/hle/service/ldr/ldr.cpp

@@ -4,7 +4,10 @@
 
 #include <memory>
 #include <fmt/format.h>
+#include <mbedtls/sha256.h>
 
+#include "common/alignment.h"
+#include "common/hex_util.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/service/ldr/ldr.h"
@@ -13,6 +16,38 @@
 
 namespace Service::LDR {
 
+namespace ErrCodes {
+enum {
+    InvalidMemoryState = 51,
+    InvalidNRO = 52,
+    InvalidNRR = 53,
+    MissingNRRHash = 54,
+    MaximumNRO = 55,
+    MaximumNRR = 56,
+    AlreadyLoaded = 57,
+    InvalidAlignment = 81,
+    InvalidSize = 82,
+    InvalidNROAddress = 84,
+    InvalidNRRAddress = 85,
+    NotInitialized = 87,
+};
+}
+
+constexpr ResultCode ERROR_INVALID_MEMORY_STATE(ErrorModule::Loader, ErrCodes::InvalidMemoryState);
+constexpr ResultCode ERROR_INVALID_NRO(ErrorModule::Loader, ErrCodes::InvalidNRO);
+constexpr ResultCode ERROR_INVALID_NRR(ErrorModule::Loader, ErrCodes::InvalidNRR);
+constexpr ResultCode ERROR_MISSING_NRR_HASH(ErrorModule::Loader, ErrCodes::MissingNRRHash);
+constexpr ResultCode ERROR_MAXIMUM_NRO(ErrorModule::Loader, ErrCodes::MaximumNRO);
+constexpr ResultCode ERROR_MAXIMUM_NRR(ErrorModule::Loader, ErrCodes::MaximumNRR);
+constexpr ResultCode ERROR_ALREADY_LOADED(ErrorModule::Loader, ErrCodes::AlreadyLoaded);
+constexpr ResultCode ERROR_INVALID_ALIGNMENT(ErrorModule::Loader, ErrCodes::InvalidAlignment);
+constexpr ResultCode ERROR_INVALID_SIZE(ErrorModule::Loader, ErrCodes::InvalidSize);
+constexpr ResultCode ERROR_INVALID_NRO_ADDRESS(ErrorModule::Loader, ErrCodes::InvalidNROAddress);
+constexpr ResultCode ERROR_INVALID_NRR_ADDRESS(ErrorModule::Loader, ErrCodes::InvalidNRRAddress);
+constexpr ResultCode ERROR_NOT_INITIALIZED(ErrorModule::Loader, ErrCodes::NotInitialized);
+
+constexpr u64 MAXIMUM_LOADED_RO = 0x40;
+
 class DebugMonitor final : public ServiceFramework<DebugMonitor> {
 public:
     explicit DebugMonitor() : ServiceFramework{"ldr:dmnt"} {
@@ -64,9 +99,9 @@ public:
         // clang-format off
         static const FunctionInfo functions[] = {
             {0, &RelocatableObject::LoadNro, "LoadNro"},
-            {1, nullptr, "UnloadNro"},
+            {1, &RelocatableObject::UnloadNro, "UnloadNro"},
             {2, &RelocatableObject::LoadNrr, "LoadNrr"},
-            {3, nullptr, "UnloadNrr"},
+            {3, &RelocatableObject::UnloadNrr, "UnloadNrr"},
             {4, &RelocatableObject::Initialize, "Initialize"},
         };
         // clang-format on
@@ -75,9 +110,123 @@ public:
     }
 
     void LoadNrr(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        rp.Skip(2, false);
+        const VAddr nrr_addr{rp.Pop<VAddr>()};
+        const u64 nrr_size{rp.Pop<u64>()};
+
+        if (!initialized) {
+            LOG_ERROR(Service_LDR, "LDR:RO not initialized before use!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_NOT_INITIALIZED);
+            return;
+        }
+
+        if (nrr.size() >= MAXIMUM_LOADED_RO) {
+            LOG_ERROR(Service_LDR, "Loading new NRR would exceed the maximum number of loaded NRRs "
+                                   "(0x40)! Failing...");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_MAXIMUM_NRR);
+            return;
+        }
+
+        // NRR Address does not fall on 0x1000 byte boundary
+        if (!Common::Is4KBAligned(nrr_addr)) {
+            LOG_ERROR(Service_LDR, "NRR Address has invalid alignment (actual {:016X})!", nrr_addr);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_ALIGNMENT);
+            return;
+        }
+
+        // NRR Size is zero or causes overflow
+        if (nrr_addr + nrr_size <= nrr_addr || nrr_size == 0 || !Common::Is4KBAligned(nrr_size)) {
+            LOG_ERROR(Service_LDR, "NRR Size is invalid! (nrr_address={:016X}, nrr_size={:016X})",
+                      nrr_addr, nrr_size);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_SIZE);
+            return;
+        }
+        // Read NRR data from memory
+        std::vector<u8> nrr_data(nrr_size);
+        Memory::ReadBlock(nrr_addr, nrr_data.data(), nrr_size);
+        NRRHeader header;
+        std::memcpy(&header, nrr_data.data(), sizeof(NRRHeader));
+
+        if (header.magic != Common::MakeMagic('N', 'R', 'R', '0')) {
+            LOG_ERROR(Service_LDR, "NRR did not have magic 'NRR0' (actual {:08X})!", header.magic);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_NRR);
+            return;
+        }
+
+        if (header.size != nrr_size) {
+            LOG_ERROR(Service_LDR,
+                      "NRR header reported size did not match LoadNrr parameter size! "
+                      "(header_size={:016X}, loadnrr_size={:016X})",
+                      header.size, nrr_size);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_SIZE);
+            return;
+        }
+
+        if (Core::CurrentProcess()->GetTitleID() != header.title_id) {
+            LOG_ERROR(Service_LDR,
+                      "Attempting to load NRR with title ID other than current process. (actual "
+                      "{:016X})!",
+                      header.title_id);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_NRR);
+            return;
+        }
+
+        std::vector<SHA256Hash> hashes;
+
+        // Copy all hashes in the NRR (specified by hash count/hash offset) into vector.
+        for (std::size_t i = header.hash_offset;
+             i < (header.hash_offset + (header.hash_count * sizeof(SHA256Hash))); i += 8) {
+            SHA256Hash hash;
+            std::memcpy(hash.data(), nrr_data.data() + i, sizeof(SHA256Hash));
+            hashes.emplace_back(hash);
+        }
+
+        nrr.insert_or_assign(nrr_addr, std::move(hashes));
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
+    void UnloadNrr(Kernel::HLERequestContext& ctx) {
+        if (!initialized) {
+            LOG_ERROR(Service_LDR, "LDR:RO not initialized before use!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_NOT_INITIALIZED);
+            return;
+        }
+
+        IPC::RequestParser rp{ctx};
+        rp.Skip(2, false);
+        const auto nrr_addr{rp.Pop<VAddr>()};
+
+        if (!Common::Is4KBAligned(nrr_addr)) {
+            LOG_ERROR(Service_LDR, "NRR Address has invalid alignment (actual {:016X})!", nrr_addr);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_ALIGNMENT);
+            return;
+        }
+
+        const auto iter = nrr.find(nrr_addr);
+        if (iter == nrr.end()) {
+            LOG_ERROR(Service_LDR,
+                      "Attempting to unload NRR which has not been loaded! (addr={:016X})",
+                      nrr_addr);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_NRR_ADDRESS);
+            return;
+        }
+
+        nrr.erase(iter);
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
-        LOG_WARNING(Service_LDR, "(STUBBED) called");
     }
 
     void LoadNro(Kernel::HLERequestContext& ctx) {
@@ -88,33 +237,253 @@ public:
         const VAddr bss_addr{rp.Pop<VAddr>()};
         const u64 bss_size{rp.Pop<u64>()};
 
+        if (!initialized) {
+            LOG_ERROR(Service_LDR, "LDR:RO not initialized before use!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_NOT_INITIALIZED);
+            return;
+        }
+
+        if (nro.size() >= MAXIMUM_LOADED_RO) {
+            LOG_ERROR(Service_LDR, "Loading new NRO would exceed the maximum number of loaded NROs "
+                                   "(0x40)! Failing...");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_MAXIMUM_NRO);
+            return;
+        }
+
+        // NRO Address does not fall on 0x1000 byte boundary
+        if (!Common::Is4KBAligned(nro_addr)) {
+            LOG_ERROR(Service_LDR, "NRO Address has invalid alignment (actual {:016X})!", nro_addr);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_ALIGNMENT);
+            return;
+        }
+
+        // NRO Size or BSS Size is zero or causes overflow
+        const auto nro_size_valid =
+            nro_size != 0 && nro_addr + nro_size > nro_addr && Common::Is4KBAligned(nro_size);
+        const auto bss_size_valid =
+            nro_size + bss_size >= nro_size && (bss_size == 0 || bss_addr + bss_size > bss_addr);
+
+        if (!nro_size_valid || !bss_size_valid) {
+            LOG_ERROR(Service_LDR,
+                      "NRO Size or BSS Size is invalid! (nro_address={:016X}, nro_size={:016X}, "
+                      "bss_address={:016X}, bss_size={:016X})",
+                      nro_addr, nro_size, bss_addr, bss_size);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_SIZE);
+            return;
+        }
+
         // Read NRO data from memory
         std::vector<u8> nro_data(nro_size);
         Memory::ReadBlock(nro_addr, nro_data.data(), nro_size);
 
-        // Load NRO as new executable module
-        const VAddr addr{*Core::CurrentProcess()->VMManager().FindFreeRegion(nro_size + bss_size)};
-        Loader::AppLoader_NRO::LoadNro(nro_data, fmt::format("nro-{:08x}", addr), addr);
+        SHA256Hash hash{};
+        mbedtls_sha256(nro_data.data(), nro_data.size(), hash.data(), 0);
 
-        // TODO(bunnei): This is an incomplete implementation. It was tested with Super Mario Party.
-        // It is currently missing:
-        // - Signature checks with LoadNRR
-        // - Checking if a module has already been loaded
-        // - Using/validating BSS, etc. params (these are used from NRO header instead)
-        // - Error checking
-        // - ...Probably other things
+        // NRO Hash is already loaded
+        if (std::any_of(nro.begin(), nro.end(), [&hash](const std::pair<VAddr, NROInfo>& info) {
+                return info.second.hash == hash;
+            })) {
+            LOG_ERROR(Service_LDR, "NRO is already loaded!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_ALREADY_LOADED);
+            return;
+        }
+
+        // NRO Hash is not in any loaded NRR
+        if (!IsValidNROHash(hash)) {
+            LOG_ERROR(Service_LDR,
+                      "NRO hash is not present in any currently loaded NRRs (hash={})!",
+                      Common::HexArrayToString(hash));
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_MISSING_NRR_HASH);
+            return;
+        }
+
+        NROHeader header;
+        std::memcpy(&header, nro_data.data(), sizeof(NROHeader));
+
+        if (!IsValidNRO(header, nro_size, bss_size)) {
+            LOG_ERROR(Service_LDR, "NRO was invalid!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_NRO);
+            return;
+        }
+
+        // Load NRO as new executable module
+        auto* process = Core::CurrentProcess();
+        auto& vm_manager = process->VMManager();
+        auto map_address = vm_manager.FindFreeRegion(nro_size + bss_size);
+
+        if (!map_address.Succeeded() ||
+            *map_address + nro_size + bss_size > vm_manager.GetAddressSpaceEndAddress()) {
+
+            LOG_ERROR(Service_LDR,
+                      "General error while allocation memory or no available memory to allocate!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_MEMORY_STATE);
+            return;
+        }
+
+        ASSERT(process->MirrorMemory(*map_address, nro_addr, nro_size,
+                                     Kernel::MemoryState::ModuleCodeStatic) == RESULT_SUCCESS);
+        ASSERT(process->UnmapMemory(nro_addr, 0, nro_size) == RESULT_SUCCESS);
+
+        if (bss_size > 0) {
+            ASSERT(process->MirrorMemory(*map_address + nro_size, bss_addr, bss_size,
+                                         Kernel::MemoryState::ModuleCodeStatic) == RESULT_SUCCESS);
+            ASSERT(process->UnmapMemory(bss_addr, 0, bss_size) == RESULT_SUCCESS);
+        }
+
+        vm_manager.ReprotectRange(*map_address, header.text_size,
+                                  Kernel::VMAPermission::ReadExecute);
+        vm_manager.ReprotectRange(*map_address + header.ro_offset, header.ro_size,
+                                  Kernel::VMAPermission::Read);
+        vm_manager.ReprotectRange(*map_address + header.rw_offset, header.rw_size,
+                                  Kernel::VMAPermission::ReadWrite);
+
+        Core::System::GetInstance().ArmInterface(0).ClearInstructionCache();
+        Core::System::GetInstance().ArmInterface(1).ClearInstructionCache();
+        Core::System::GetInstance().ArmInterface(2).ClearInstructionCache();
+        Core::System::GetInstance().ArmInterface(3).ClearInstructionCache();
+
+        nro.insert_or_assign(*map_address, NROInfo{hash, nro_size + bss_size});
 
         IPC::ResponseBuilder rb{ctx, 4};
         rb.Push(RESULT_SUCCESS);
-        rb.Push(addr);
-        LOG_WARNING(Service_LDR, "(STUBBED) called");
+        rb.Push(*map_address);
+    }
+
+    void UnloadNro(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        rp.Skip(2, false);
+        const VAddr mapped_addr{rp.PopRaw<VAddr>()};
+        const VAddr heap_addr{rp.PopRaw<VAddr>()};
+
+        if (!initialized) {
+            LOG_ERROR(Service_LDR, "LDR:RO not initialized before use!");
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_NOT_INITIALIZED);
+            return;
+        }
+
+        if (!Common::Is4KBAligned(mapped_addr) || !Common::Is4KBAligned(heap_addr)) {
+            LOG_ERROR(Service_LDR,
+                      "NRO/BSS Address has invalid alignment (actual nro_addr={:016X}, "
+                      "bss_addr={:016X})!",
+                      mapped_addr, heap_addr);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_ALIGNMENT);
+            return;
+        }
+
+        const auto iter = nro.find(mapped_addr);
+        if (iter == nro.end()) {
+            LOG_ERROR(Service_LDR,
+                      "The NRO attempting to unmap was not mapped or has an invalid address "
+                      "(actual {:016X})!",
+                      mapped_addr);
+            IPC::ResponseBuilder rb{ctx, 2};
+            rb.Push(ERROR_INVALID_NRO_ADDRESS);
+            return;
+        }
+
+        auto* process = Core::CurrentProcess();
+        auto& vm_manager = process->VMManager();
+        const auto& nro_size = iter->second.size;
+
+        ASSERT(process->MirrorMemory(heap_addr, mapped_addr, nro_size,
+                                     Kernel::MemoryState::ModuleCodeStatic) == RESULT_SUCCESS);
+        ASSERT(process->UnmapMemory(mapped_addr, 0, nro_size) == RESULT_SUCCESS);
+
+        Core::System::GetInstance().ArmInterface(0).ClearInstructionCache();
+        Core::System::GetInstance().ArmInterface(1).ClearInstructionCache();
+        Core::System::GetInstance().ArmInterface(2).ClearInstructionCache();
+        Core::System::GetInstance().ArmInterface(3).ClearInstructionCache();
+
+        nro.erase(iter);
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
     }
 
     void Initialize(Kernel::HLERequestContext& ctx) {
+        initialized = true;
+
         IPC::ResponseBuilder rb{ctx, 2};
         rb.Push(RESULT_SUCCESS);
         LOG_WARNING(Service_LDR, "(STUBBED) called");
     }
+
+private:
+    using SHA256Hash = std::array<u8, 0x20>;
+
+    struct NROHeader {
+        u32_le entrypoint_insn;
+        u32_le mod_offset;
+        INSERT_PADDING_WORDS(2);
+        u32_le magic;
+        INSERT_PADDING_WORDS(1);
+        u32_le nro_size;
+        INSERT_PADDING_WORDS(1);
+        u32_le text_offset;
+        u32_le text_size;
+        u32_le ro_offset;
+        u32_le ro_size;
+        u32_le rw_offset;
+        u32_le rw_size;
+        u32_le bss_size;
+        INSERT_PADDING_WORDS(1);
+        std::array<u8, 0x20> build_id;
+        INSERT_PADDING_BYTES(0x20);
+    };
+    static_assert(sizeof(NROHeader) == 0x80, "NROHeader has invalid size.");
+
+    struct NRRHeader {
+        u32_le magic;
+        INSERT_PADDING_BYTES(0x1C);
+        u64_le title_id_mask;
+        u64_le title_id_pattern;
+        std::array<u8, 0x100> modulus;
+        std::array<u8, 0x100> signature_1;
+        std::array<u8, 0x100> signature_2;
+        u64_le title_id;
+        u32_le size;
+        INSERT_PADDING_BYTES(4);
+        u32_le hash_offset;
+        u32_le hash_count;
+        INSERT_PADDING_BYTES(8);
+    };
+    static_assert(sizeof(NRRHeader) == 0x350, "NRRHeader has incorrect size.");
+
+    struct NROInfo {
+        SHA256Hash hash;
+        u64 size;
+    };
+
+    bool initialized = false;
+
+    std::map<VAddr, NROInfo> nro;
+    std::map<VAddr, std::vector<SHA256Hash>> nrr;
+
+    bool IsValidNROHash(const SHA256Hash& hash) {
+        return std::any_of(
+            nrr.begin(), nrr.end(), [&hash](const std::pair<VAddr, std::vector<SHA256Hash>>& p) {
+                return std::find(p.second.begin(), p.second.end(), hash) != p.second.end();
+            });
+    }
+
+    static bool IsValidNRO(const NROHeader& header, u64 nro_size, u64 bss_size) {
+        return header.magic == Common::MakeMagic('N', 'R', 'O', '0') &&
+               header.nro_size == nro_size && header.bss_size == bss_size &&
+               header.ro_offset == header.text_offset + header.text_size &&
+               header.rw_offset == header.ro_offset + header.ro_size &&
+               nro_size == header.rw_offset + header.rw_size &&
+               Common::Is4KBAligned(header.text_size) && Common::Is4KBAligned(header.ro_size) &&
+               Common::Is4KBAligned(header.rw_size);
+    }
 };
 
 void InstallInterfaces(SM::ServiceManager& sm) {

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

@@ -351,6 +351,14 @@ void PL_U::GetSharedFontInOrderOfPriority(Kernel::HLERequestContext& ctx) {
         font_sizes.push_back(region.size);
     }
 
+    // Resize buffers if game requests smaller size output.
+    font_codes.resize(
+        std::min<std::size_t>(font_codes.size(), ctx.GetWriteBufferSize(0) / sizeof(u32)));
+    font_offsets.resize(
+        std::min<std::size_t>(font_offsets.size(), ctx.GetWriteBufferSize(1) / sizeof(u32)));
+    font_sizes.resize(
+        std::min<std::size_t>(font_sizes.size(), ctx.GetWriteBufferSize(2) / sizeof(u32)));
+
     ctx.WriteBuffer(font_codes, 0);
     ctx.WriteBuffer(font_offsets, 1);
     ctx.WriteBuffer(font_sizes, 2);

src/core/hle/service/time/interface.cpp

@@ -23,7 +23,8 @@ Time::Time(std::shared_ptr<Module> time, const char* name)
         {300, nullptr, "CalculateMonotonicSystemClockBaseTimePoint"},
         {400, &Time::GetClockSnapshot, "GetClockSnapshot"},
         {401, nullptr, "GetClockSnapshotFromSystemClockContext"},
-        {500, nullptr, "CalculateStandardUserSystemClockDifferenceByUser"},
+        {500, &Time::CalculateStandardUserSystemClockDifferenceByUser,
+         "CalculateStandardUserSystemClockDifferenceByUser"},
         {501, nullptr, "CalculateSpanBetween"},
     };
     RegisterHandlers(functions);

src/core/hle/service/time/time.cpp

@@ -299,6 +299,21 @@ void Module::Interface::GetClockSnapshot(Kernel::HLERequestContext& ctx) {
     ctx.WriteBuffer(&clock_snapshot, sizeof(ClockSnapshot));
 }
 
+void Module::Interface::CalculateStandardUserSystemClockDifferenceByUser(
+    Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_Time, "called");
+
+    IPC::RequestParser rp{ctx};
+    const auto snapshot_a = rp.PopRaw<ClockSnapshot>();
+    const auto snapshot_b = rp.PopRaw<ClockSnapshot>();
+    const u64 difference =
+        snapshot_b.user_clock_context.offset - snapshot_a.user_clock_context.offset;
+
+    IPC::ResponseBuilder rb{ctx, 4};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushRaw<u64>(difference);
+}
+
 Module::Interface::Interface(std::shared_ptr<Module> time, const char* name)
     : ServiceFramework(name), time(std::move(time)) {}
 

src/core/hle/service/time/time.h

@@ -84,6 +84,7 @@ public:
         void GetTimeZoneService(Kernel::HLERequestContext& ctx);
         void GetStandardLocalSystemClock(Kernel::HLERequestContext& ctx);
         void GetClockSnapshot(Kernel::HLERequestContext& ctx);
+        void CalculateStandardUserSystemClockDifferenceByUser(Kernel::HLERequestContext& ctx);
 
     protected:
         std::shared_ptr<Module> time;

src/core/hle/service/vi/vi.cpp

@@ -237,6 +237,22 @@ private:
     Data data{};
 };
 
+/// Represents a parcel containing one int '0' as its data
+/// Used by DetachBuffer and Disconnect
+class IGBPEmptyResponseParcel : public Parcel {
+protected:
+    void SerializeData() override {
+        Write(data);
+    }
+
+private:
+    struct Data {
+        u32_le unk_0;
+    };
+
+    Data data{};
+};
+
 class IGBPSetPreallocatedBufferRequestParcel : public Parcel {
 public:
     explicit IGBPSetPreallocatedBufferRequestParcel(std::vector<u8> buffer)
@@ -554,6 +570,12 @@ private:
             ctx.WriteBuffer(response.Serialize());
         } else if (transaction == TransactionId::CancelBuffer) {
             LOG_CRITICAL(Service_VI, "(STUBBED) called, transaction=CancelBuffer");
+        } else if (transaction == TransactionId::Disconnect ||
+                   transaction == TransactionId::DetachBuffer) {
+            const auto buffer = ctx.ReadBuffer();
+
+            IGBPEmptyResponseParcel response{};
+            ctx.WriteBuffer(response.Serialize());
         } else {
             ASSERT_MSG(false, "Unimplemented");
         }

src/core/loader/nro.cpp

@@ -170,17 +170,20 @@ static constexpr u32 PageAlignSize(u32 size) {
                     arg_data.size());
     }
 
-    // Read MOD header
-    ModHeader mod_header{};
     // Default .bss to NRO header bss size if MOD0 section doesn't exist
     u32 bss_size{PageAlignSize(nro_header.bss_size)};
+
+    // Read MOD header
+    ModHeader mod_header{};
     std::memcpy(&mod_header, program_image.data() + nro_header.module_header_offset,
                 sizeof(ModHeader));
+
     const bool has_mod_header{mod_header.magic == Common::MakeMagic('M', 'O', 'D', '0')};
     if (has_mod_header) {
         // Resize program image to include .bss section and page align each section
         bss_size = PageAlignSize(mod_header.bss_end_offset - mod_header.bss_start_offset);
     }
+
     codeset.DataSegment().size += bss_size;
     program_image.resize(static_cast<u32>(program_image.size()) + bss_size);
 

src/video_core/command_processor.cpp

@@ -34,9 +34,6 @@ MICROPROFILE_DEFINE(ProcessCommandLists, "GPU", "Execute command buffer", MP_RGB
 void GPU::ProcessCommandLists(const std::vector<CommandListHeader>& commands) {
     MICROPROFILE_SCOPE(ProcessCommandLists);
 
-    // On entering GPU code, assume all memory may be touched by the ARM core.
-    maxwell_3d->dirty_flags.OnMemoryWrite();
-
     auto WriteReg = [this](u32 method, u32 subchannel, u32 value, u32 remaining_params) {
         LOG_TRACE(HW_GPU,
                   "Processing method {:08X} on subchannel {} value "

src/video_core/engines/fermi_2d.cpp

@@ -2,10 +2,8 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "core/core.h"
 #include "core/memory.h"
 #include "video_core/engines/fermi_2d.h"
-#include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/textures/decoders.h"
 
@@ -49,9 +47,6 @@ void Fermi2D::HandleSurfaceCopy() {
     u32 dst_bytes_per_pixel = RenderTargetBytesPerPixel(regs.dst.format);
 
     if (!rasterizer.AccelerateSurfaceCopy(regs.src, regs.dst)) {
-        // All copies here update the main memory, so mark all rasterizer states as invalid.
-        Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
-
         rasterizer.FlushRegion(source_cpu, src_bytes_per_pixel * regs.src.width * regs.src.height);
         // We have to invalidate the destination region to evict any outdated surfaces from the
         // cache. We do this before actually writing the new data because the destination address

src/video_core/engines/kepler_memory.cpp

@@ -3,10 +3,8 @@
 // Refer to the license.txt file included.
 
 #include "common/logging/log.h"
-#include "core/core.h"
 #include "core/memory.h"
 #include "video_core/engines/kepler_memory.h"
-#include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
 
 namespace Tegra::Engines {
@@ -49,7 +47,6 @@ void KeplerMemory::ProcessData(u32 data) {
     rasterizer.InvalidateRegion(dest_address, sizeof(u32));
 
     Memory::Write32(dest_address, data);
-    Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
 
     state.write_offset++;
 }

src/video_core/engines/maxwell_3d.cpp

@@ -34,8 +34,8 @@ void Maxwell3D::InitializeRegisterDefaults() {
     // Depth range near/far is not always set, but is expected to be the default 0.0f, 1.0f. This is
     // needed for ARMS.
     for (std::size_t viewport{}; viewport < Regs::NumViewports; ++viewport) {
-        regs.viewport[viewport].depth_range_near = 0.0f;
-        regs.viewport[viewport].depth_range_far = 1.0f;
+        regs.viewports[viewport].depth_range_near = 0.0f;
+        regs.viewports[viewport].depth_range_far = 1.0f;
     }
     // Doom and Bomberman seems to use the uninitialized registers and just enable blend
     // so initialize blend registers with sane values
@@ -66,6 +66,9 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.stencil_back_func_func = Regs::ComparisonOp::Always;
     regs.stencil_back_func_mask = 0xFFFFFFFF;
     regs.stencil_back_mask = 0xFFFFFFFF;
+    // TODO(Rodrigo): Most games do not set a point size. I think this is a case of a
+    // register carrying a default value. Assume it's OpenGL's default (1).
+    regs.point_size = 1.0f;
 }
 
 void Maxwell3D::CallMacroMethod(u32 method, std::vector<u32> parameters) {
@@ -123,24 +126,10 @@ void Maxwell3D::WriteReg(u32 method, u32 value, u32 remaining_params) {
 
     if (regs.reg_array[method] != value) {
         regs.reg_array[method] = value;
-        // Vertex format
         if (method >= MAXWELL3D_REG_INDEX(vertex_attrib_format) &&
             method < MAXWELL3D_REG_INDEX(vertex_attrib_format) + regs.vertex_attrib_format.size()) {
             dirty_flags.vertex_attrib_format = true;
         }
-
-        // Vertex buffer
-        if (method >= MAXWELL3D_REG_INDEX(vertex_array) &&
-            method < MAXWELL3D_REG_INDEX(vertex_array) + 4 * 32) {
-            dirty_flags.vertex_array |= 1u << ((method - MAXWELL3D_REG_INDEX(vertex_array)) >> 2);
-        } else if (method >= MAXWELL3D_REG_INDEX(vertex_array_limit) &&
-                   method < MAXWELL3D_REG_INDEX(vertex_array_limit) + 2 * 32) {
-            dirty_flags.vertex_array |=
-                1u << ((method - MAXWELL3D_REG_INDEX(vertex_array_limit)) >> 1);
-        } else if (method >= MAXWELL3D_REG_INDEX(instanced_arrays) &&
-                   method < MAXWELL3D_REG_INDEX(instanced_arrays) + 32) {
-            dirty_flags.vertex_array |= 1u << (method - MAXWELL3D_REG_INDEX(instanced_arrays));
-        }
     }
 
     switch (method) {
@@ -272,7 +261,6 @@ void Maxwell3D::ProcessQueryGet() {
             query_result.timestamp = CoreTiming::GetTicks();
             Memory::WriteBlock(*address, &query_result, sizeof(query_result));
         }
-        dirty_flags.OnMemoryWrite();
         break;
     }
     default:
@@ -349,7 +337,6 @@ void Maxwell3D::ProcessCBData(u32 value) {
         memory_manager.GpuToCpuAddress(buffer_address + regs.const_buffer.cb_pos);
 
     Memory::Write32(*address, value);
-    dirty_flags.OnMemoryWrite();
 
     // Increment the current buffer position.
     regs.const_buffer.cb_pos = regs.const_buffer.cb_pos + 4;

src/video_core/engines/maxwell_3d.h

@@ -480,6 +480,67 @@ public:
             };
         };
 
+        struct ViewportTransform {
+            f32 scale_x;
+            f32 scale_y;
+            f32 scale_z;
+            f32 translate_x;
+            f32 translate_y;
+            f32 translate_z;
+            INSERT_PADDING_WORDS(2);
+
+            MathUtil::Rectangle<s32> GetRect() const {
+                return {
+                    GetX(),               // left
+                    GetY() + GetHeight(), // top
+                    GetX() + GetWidth(),  // right
+                    GetY()                // bottom
+                };
+            };
+
+            s32 GetX() const {
+                return static_cast<s32>(std::max(0.0f, translate_x - std::fabs(scale_x)));
+            }
+
+            s32 GetY() const {
+                return static_cast<s32>(std::max(0.0f, translate_y - std::fabs(scale_y)));
+            }
+
+            s32 GetWidth() const {
+                return static_cast<s32>(translate_x + std::fabs(scale_x)) - GetX();
+            }
+
+            s32 GetHeight() const {
+                return static_cast<s32>(translate_y + std::fabs(scale_y)) - GetY();
+            }
+        };
+
+        struct ScissorTest {
+            u32 enable;
+            union {
+                BitField<0, 16, u32> min_x;
+                BitField<16, 16, u32> max_x;
+            };
+            union {
+                BitField<0, 16, u32> min_y;
+                BitField<16, 16, u32> max_y;
+            };
+            u32 fill;
+        };
+
+        struct ViewPort {
+            union {
+                BitField<0, 16, u32> x;
+                BitField<16, 16, u32> width;
+            };
+            union {
+                BitField<0, 16, u32> y;
+                BitField<16, 16, u32> height;
+            };
+            float depth_range_near;
+            float depth_range_far;
+        };
+
         bool IsShaderConfigEnabled(std::size_t index) const {
             // The VertexB is always enabled.
             if (index == static_cast<std::size_t>(Regs::ShaderProgram::VertexB)) {
@@ -505,55 +566,11 @@ public:
 
                 INSERT_PADDING_WORDS(0x2E);
 
-                RenderTargetConfig rt[NumRenderTargets];
+                std::array<RenderTargetConfig, NumRenderTargets> rt;
 
-                struct {
-                    f32 scale_x;
-                    f32 scale_y;
-                    f32 scale_z;
-                    f32 translate_x;
-                    f32 translate_y;
-                    f32 translate_z;
-                    INSERT_PADDING_WORDS(2);
+                std::array<ViewportTransform, NumViewports> viewport_transform;
 
-                    MathUtil::Rectangle<s32> GetRect() const {
-                        return {
-                            GetX(),               // left
-                            GetY() + GetHeight(), // top
-                            GetX() + GetWidth(),  // right
-                            GetY()                // bottom
-                        };
-                    };
-
-                    s32 GetX() const {
-                        return static_cast<s32>(std::max(0.0f, translate_x - std::fabs(scale_x)));
-                    }
-
-                    s32 GetY() const {
-                        return static_cast<s32>(std::max(0.0f, translate_y - std::fabs(scale_y)));
-                    }
-
-                    s32 GetWidth() const {
-                        return static_cast<s32>(translate_x + std::fabs(scale_x)) - GetX();
-                    }
-
-                    s32 GetHeight() const {
-                        return static_cast<s32>(translate_y + std::fabs(scale_y)) - GetY();
-                    }
-                } viewport_transform[NumViewports];
-
-                struct {
-                    union {
-                        BitField<0, 16, u32> x;
-                        BitField<16, 16, u32> width;
-                    };
-                    union {
-                        BitField<0, 16, u32> y;
-                        BitField<16, 16, u32> height;
-                    };
-                    float depth_range_near;
-                    float depth_range_far;
-                } viewport[NumViewports];
+                std::array<ViewPort, NumViewports> viewports;
 
                 INSERT_PADDING_WORDS(0x1D);
 
@@ -571,19 +588,9 @@ public:
 
                 INSERT_PADDING_WORDS(0x17);
 
-                struct {
-                    u32 enable;
-                    union {
-                        BitField<0, 16, u32> min_x;
-                        BitField<16, 16, u32> max_x;
-                    };
-                    union {
-                        BitField<0, 16, u32> min_y;
-                        BitField<16, 16, u32> max_y;
-                    };
-                } scissor_test;
+                std::array<ScissorTest, NumViewports> scissor_test;
 
-                INSERT_PADDING_WORDS(0x52);
+                INSERT_PADDING_WORDS(0x15);
 
                 s32 stencil_back_func_ref;
                 u32 stencil_back_mask;
@@ -700,7 +707,9 @@ public:
                 u32 stencil_front_func_mask;
                 u32 stencil_front_mask;
 
-                INSERT_PADDING_WORDS(0x3);
+                INSERT_PADDING_WORDS(0x2);
+
+                u32 frag_color_clamp;
 
                 union {
                     BitField<4, 1, u32> triangle_rast_flip;
@@ -718,7 +727,12 @@ public:
 
                 u32 zeta_enable;
 
-                INSERT_PADDING_WORDS(0x8);
+                union {
+                    BitField<0, 1, u32> alpha_to_coverage;
+                    BitField<4, 1, u32> alpha_to_one;
+                } multisample_control;
+
+                INSERT_PADDING_WORDS(0x7);
 
                 struct {
                     u32 tsc_address_high;
@@ -1014,11 +1028,6 @@ public:
 
     struct DirtyFlags {
         bool vertex_attrib_format = true;
-        u32 vertex_array = 0xFFFFFFFF;
-
-        void OnMemoryWrite() {
-            vertex_array = 0xFFFFFFFF;
-        }
     };
 
     DirtyFlags dirty_flags;
@@ -1105,8 +1114,8 @@ private:
 ASSERT_REG_POSITION(macros, 0x45);
 ASSERT_REG_POSITION(tfb_enabled, 0x1D1);
 ASSERT_REG_POSITION(rt, 0x200);
-ASSERT_REG_POSITION(viewport_transform[0], 0x280);
-ASSERT_REG_POSITION(viewport, 0x300);
+ASSERT_REG_POSITION(viewport_transform, 0x280);
+ASSERT_REG_POSITION(viewports, 0x300);
 ASSERT_REG_POSITION(vertex_buffer, 0x35D);
 ASSERT_REG_POSITION(clear_color[0], 0x360);
 ASSERT_REG_POSITION(clear_depth, 0x364);
@@ -1141,10 +1150,12 @@ ASSERT_REG_POSITION(stencil_front_func_func, 0x4E4);
 ASSERT_REG_POSITION(stencil_front_func_ref, 0x4E5);
 ASSERT_REG_POSITION(stencil_front_func_mask, 0x4E6);
 ASSERT_REG_POSITION(stencil_front_mask, 0x4E7);
+ASSERT_REG_POSITION(frag_color_clamp, 0x4EA);
 ASSERT_REG_POSITION(screen_y_control, 0x4EB);
 ASSERT_REG_POSITION(vb_element_base, 0x50D);
 ASSERT_REG_POSITION(point_size, 0x546);
 ASSERT_REG_POSITION(zeta_enable, 0x54E);
+ASSERT_REG_POSITION(multisample_control, 0x54F);
 ASSERT_REG_POSITION(tsc, 0x557);
 ASSERT_REG_POSITION(tic, 0x55D);
 ASSERT_REG_POSITION(stencil_two_side_enable, 0x565);

src/video_core/engines/maxwell_dma.cpp

@@ -2,9 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
-#include "core/core.h"
 #include "core/memory.h"
-#include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/maxwell_dma.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/textures/decoders.h"
@@ -56,9 +54,6 @@ void MaxwellDMA::HandleCopy() {
         return;
     }
 
-    // All copies here update the main memory, so mark all rasterizer states as invalid.
-    Core::System::GetInstance().GPU().Maxwell3D().dirty_flags.OnMemoryWrite();
-
     if (regs.exec.is_dst_linear && regs.exec.is_src_linear) {
         // When the enable_2d bit is disabled, the copy is performed as if we were copying a 1D
         // buffer of length `x_count`, otherwise we copy a 2D image of dimensions (x_count,

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -76,7 +76,7 @@ std::tuple<u8*, GLintptr> OGLBufferCache::ReserveMemory(std::size_t size, std::s
     return std::make_tuple(uploaded_ptr, uploaded_offset);
 }
 
-bool OGLBufferCache::Map(std::size_t max_size) {
+void OGLBufferCache::Map(std::size_t max_size) {
     bool invalidate;
     std::tie(buffer_ptr, buffer_offset_base, invalidate) =
         stream_buffer.Map(static_cast<GLsizeiptr>(max_size), 4);
@@ -85,7 +85,6 @@ bool OGLBufferCache::Map(std::size_t max_size) {
     if (invalidate) {
         InvalidateAll();
     }
-    return invalidate;
 }
 
 void OGLBufferCache::Unmap() {

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -50,7 +50,7 @@ public:
     /// Reserves memory to be used by host's CPU. Returns mapped address and offset.
     std::tuple<u8*, GLintptr> ReserveMemory(std::size_t size, std::size_t alignment = 4);
 
-    bool Map(std::size_t max_size);
+    void Map(std::size_t max_size);
     void Unmap();
 
     GLuint GetHandle() const;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -107,8 +107,6 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo
 
     ASSERT_MSG(has_ARB_separate_shader_objects, "has_ARB_separate_shader_objects is unsupported");
     OpenGLState::ApplyDefaultState();
-    // Clipping plane 0 is always enabled for PICA fixed clip plane z <= 0
-    state.clip_distance[0] = true;
 
     // Create render framebuffer
     framebuffer.Create();
@@ -183,25 +181,15 @@ void RasterizerOpenGL::SetupVertexFormat() {
     }
     state.draw.vertex_array = VAO.handle;
     state.ApplyVertexBufferState();
-
-    // Rebinding the VAO invalidates the vertex buffer bindings.
-    gpu.dirty_flags.vertex_array = 0xFFFFFFFF;
 }
 
 void RasterizerOpenGL::SetupVertexBuffer() {
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
-    const auto& regs = gpu.regs;
-
-    if (!gpu.dirty_flags.vertex_array)
-        return;
-
     MICROPROFILE_SCOPE(OpenGL_VB);
+    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    const auto& regs = gpu.regs;
 
     // Upload all guest vertex arrays sequentially to our buffer
     for (u32 index = 0; index < Maxwell::NumVertexArrays; ++index) {
-        if (~gpu.dirty_flags.vertex_array & (1u << index))
-            continue;
-
         const auto& vertex_array = regs.vertex_array[index];
         if (!vertex_array.IsEnabled())
             continue;
@@ -228,8 +216,6 @@ void RasterizerOpenGL::SetupVertexBuffer() {
 
     // Implicit set by glBindVertexBuffer. Stupid glstate handling...
     state.draw.vertex_buffer = buffer_cache.GetHandle();
-
-    gpu.dirty_flags.vertex_array = 0;
 }
 
 DrawParameters RasterizerOpenGL::SetupDraw() {
@@ -587,13 +573,15 @@ void RasterizerOpenGL::DrawArrays() {
         return;
 
     MICROPROFILE_SCOPE(OpenGL_Drawing);
-    auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     const auto& regs = gpu.regs;
 
     ScopeAcquireGLContext acquire_context{emu_window};
 
     ConfigureFramebuffers(state);
     SyncColorMask();
+    SyncFragmentColorClampState();
+    SyncMultiSampleState();
     SyncDepthTestState();
     SyncStencilTestState();
     SyncBlendState();
@@ -638,11 +626,7 @@ void RasterizerOpenGL::DrawArrays() {
     // Add space for at least 18 constant buffers
     buffer_size += Maxwell::MaxConstBuffers * (MaxConstbufferSize + uniform_buffer_alignment);
 
-    bool invalidate = buffer_cache.Map(buffer_size);
-    if (invalidate) {
-        // As all cached buffers are invalidated, we need to recheck their state.
-        gpu.dirty_flags.vertex_attrib_format = 0xFFFFFFFF;
-    }
+    buffer_cache.Map(buffer_size);
 
     SetupVertexFormat();
     SetupVertexBuffer();
@@ -658,7 +642,7 @@ void RasterizerOpenGL::DrawArrays() {
     params.DispatchDraw();
 
     // Disable scissor test
-    state.scissor.enabled = false;
+    state.viewports[0].scissor.enabled = false;
 
     accelerate_draw = AccelDraw::Disabled;
 
@@ -749,9 +733,8 @@ void RasterizerOpenGL::SamplerInfo::Create() {
     glSamplerParameteri(sampler.handle, GL_TEXTURE_COMPARE_FUNC, GL_NEVER);
 }
 
-void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::FullTextureInfo& info) {
+void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntry& config) {
     const GLuint s = sampler.handle;
-    const Tegra::Texture::TSCEntry& config = info.tsc;
     if (mag_filter != config.mag_filter) {
         mag_filter = config.mag_filter;
         glSamplerParameteri(
@@ -793,30 +776,50 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::FullTex
                             MaxwellToGL::DepthCompareFunc(depth_compare_func));
     }
 
-    if (wrap_u == Tegra::Texture::WrapMode::Border || wrap_v == Tegra::Texture::WrapMode::Border ||
-        wrap_p == Tegra::Texture::WrapMode::Border) {
-        const GLvec4 new_border_color = {{config.border_color_r, config.border_color_g,
-                                          config.border_color_b, config.border_color_a}};
-        if (border_color != new_border_color) {
-            border_color = new_border_color;
-            glSamplerParameterfv(s, GL_TEXTURE_BORDER_COLOR, border_color.data());
+    GLvec4 new_border_color;
+    if (config.srgb_conversion) {
+        new_border_color[0] = config.srgb_border_color_r / 255.0f;
+        new_border_color[1] = config.srgb_border_color_g / 255.0f;
+        new_border_color[2] = config.srgb_border_color_g / 255.0f;
+    } else {
+        new_border_color[0] = config.border_color_r;
+        new_border_color[1] = config.border_color_g;
+        new_border_color[2] = config.border_color_b;
+    }
+    new_border_color[3] = config.border_color_a;
+
+    if (border_color != new_border_color) {
+        border_color = new_border_color;
+        glSamplerParameterfv(s, GL_TEXTURE_BORDER_COLOR, border_color.data());
+    }
+
+    const float anisotropic_max = static_cast<float>(1 << config.max_anisotropy.Value());
+    if (anisotropic_max != max_anisotropic) {
+        max_anisotropic = anisotropic_max;
+        if (GLAD_GL_ARB_texture_filter_anisotropic) {
+            glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY, max_anisotropic);
+        } else if (GLAD_GL_EXT_texture_filter_anisotropic) {
+            glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY_EXT, max_anisotropic);
         }
     }
-    if (info.tic.use_header_opt_control == 0) {
-        if (GLAD_GL_ARB_texture_filter_anisotropic) {
-            glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY,
-                                static_cast<float>(1 << info.tic.max_anisotropy.Value()));
-        } else if (GLAD_GL_EXT_texture_filter_anisotropic) {
-            glSamplerParameterf(s, GL_TEXTURE_MAX_ANISOTROPY_EXT,
-                                static_cast<float>(1 << info.tic.max_anisotropy.Value()));
-        }
-        glSamplerParameterf(s, GL_TEXTURE_MIN_LOD,
-                            static_cast<float>(info.tic.res_min_mip_level.Value()));
-        glSamplerParameterf(s, GL_TEXTURE_MAX_LOD,
-                            static_cast<float>(info.tic.res_max_mip_level.Value() == 0
-                                                   ? 16
-                                                   : info.tic.res_max_mip_level.Value()));
-        glSamplerParameterf(s, GL_TEXTURE_LOD_BIAS, info.tic.mip_lod_bias.Value() / 256.f);
+    const float lod_min = static_cast<float>(config.min_lod_clamp.Value()) / 256.0f;
+    if (lod_min != min_lod) {
+        min_lod = lod_min;
+        glSamplerParameterf(s, GL_TEXTURE_MIN_LOD, min_lod);
+    }
+
+    const float lod_max = static_cast<float>(config.max_lod_clamp.Value()) / 256.0f;
+    if (lod_max != max_lod) {
+        max_lod = lod_max;
+        glSamplerParameterf(s, GL_TEXTURE_MAX_LOD, max_lod);
+    }
+    const u32 bias = config.mip_lod_bias.Value();
+    // Sign extend the 13-bit value.
+    const u32 mask = 1U << (13 - 1);
+    const float bias_lod = static_cast<s32>((bias ^ mask) - mask) / 256.f;
+    if (lod_bias != bias_lod) {
+        lod_bias = bias_lod;
+        glSamplerParameterf(s, GL_TEXTURE_LOD_BIAS, lod_bias);
     }
 }
 
@@ -915,7 +918,7 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
             continue;
         }
 
-        texture_samplers[current_bindpoint].SyncWithConfig(texture);
+        texture_samplers[current_bindpoint].SyncWithConfig(texture.tsc);
         Surface surface = res_cache.GetTextureSurface(texture, entry);
         if (surface != nullptr) {
             state.texture_units[current_bindpoint].texture = surface->Texture().handle;
@@ -939,15 +942,15 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
 
 void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-    for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
+    for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumViewports; i++) {
         const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[i].GetRect()};
         auto& viewport = current_state.viewports[i];
         viewport.x = viewport_rect.left;
         viewport.y = viewport_rect.bottom;
         viewport.width = static_cast<GLfloat>(viewport_rect.GetWidth());
         viewport.height = static_cast<GLfloat>(viewport_rect.GetHeight());
-        viewport.depth_range_far = regs.viewport[i].depth_range_far;
-        viewport.depth_range_near = regs.viewport[i].depth_range_near;
+        viewport.depth_range_far = regs.viewports[i].depth_range_far;
+        viewport.depth_range_near = regs.viewports[i].depth_range_near;
     }
 }
 
@@ -1038,7 +1041,9 @@ void RasterizerOpenGL::SyncStencilTestState() {
 
 void RasterizerOpenGL::SyncColorMask() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-    for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
+    const std::size_t count =
+        regs.independent_blend_enable ? Tegra::Engines::Maxwell3D::Regs::NumRenderTargets : 1;
+    for (std::size_t i = 0; i < count; i++) {
         const auto& source = regs.color_mask[regs.color_mask_common ? 0 : i];
         auto& dest = state.color_mask[i];
         dest.red_enabled = (source.R == 0) ? GL_FALSE : GL_TRUE;
@@ -1048,6 +1053,17 @@ void RasterizerOpenGL::SyncColorMask() {
     }
 }
 
+void RasterizerOpenGL::SyncMultiSampleState() {
+    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    state.multisample_control.alpha_to_coverage = regs.multisample_control.alpha_to_coverage != 0;
+    state.multisample_control.alpha_to_one = regs.multisample_control.alpha_to_one != 0;
+}
+
+void RasterizerOpenGL::SyncFragmentColorClampState() {
+    const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
+    state.fragment_color_clamp.enabled = regs.frag_color_clamp != 0;
+}
+
 void RasterizerOpenGL::SyncBlendState() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
 
@@ -1059,43 +1075,40 @@ void RasterizerOpenGL::SyncBlendState() {
     state.independant_blend.enabled = regs.independent_blend_enable;
     if (!state.independant_blend.enabled) {
         auto& blend = state.blend[0];
-        blend.enabled = regs.blend.enable[0] != 0;
-        blend.separate_alpha = regs.blend.separate_alpha;
-        blend.rgb_equation = MaxwellToGL::BlendEquation(regs.blend.equation_rgb);
-        blend.src_rgb_func = MaxwellToGL::BlendFunc(regs.blend.factor_source_rgb);
-        blend.dst_rgb_func = MaxwellToGL::BlendFunc(regs.blend.factor_dest_rgb);
-        if (blend.separate_alpha) {
-            blend.a_equation = MaxwellToGL::BlendEquation(regs.blend.equation_a);
-            blend.src_a_func = MaxwellToGL::BlendFunc(regs.blend.factor_source_a);
-            blend.dst_a_func = MaxwellToGL::BlendFunc(regs.blend.factor_dest_a);
+        const auto& src = regs.blend;
+        blend.enabled = src.enable[0] != 0;
+        if (blend.enabled) {
+            blend.rgb_equation = MaxwellToGL::BlendEquation(src.equation_rgb);
+            blend.src_rgb_func = MaxwellToGL::BlendFunc(src.factor_source_rgb);
+            blend.dst_rgb_func = MaxwellToGL::BlendFunc(src.factor_dest_rgb);
+            blend.a_equation = MaxwellToGL::BlendEquation(src.equation_a);
+            blend.src_a_func = MaxwellToGL::BlendFunc(src.factor_source_a);
+            blend.dst_a_func = MaxwellToGL::BlendFunc(src.factor_dest_a);
         }
-        for (size_t i = 1; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
+        for (std::size_t i = 1; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
             state.blend[i].enabled = false;
         }
         return;
     }
 
-    for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
+    for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
         auto& blend = state.blend[i];
+        const auto& src = regs.independent_blend[i];
         blend.enabled = regs.blend.enable[i] != 0;
         if (!blend.enabled)
             continue;
-        blend.separate_alpha = regs.independent_blend[i].separate_alpha;
-        blend.rgb_equation = MaxwellToGL::BlendEquation(regs.independent_blend[i].equation_rgb);
-        blend.src_rgb_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_source_rgb);
-        blend.dst_rgb_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_dest_rgb);
-        if (blend.separate_alpha) {
-            blend.a_equation = MaxwellToGL::BlendEquation(regs.independent_blend[i].equation_a);
-            blend.src_a_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_source_a);
-            blend.dst_a_func = MaxwellToGL::BlendFunc(regs.independent_blend[i].factor_dest_a);
-        }
+        blend.rgb_equation = MaxwellToGL::BlendEquation(src.equation_rgb);
+        blend.src_rgb_func = MaxwellToGL::BlendFunc(src.factor_source_rgb);
+        blend.dst_rgb_func = MaxwellToGL::BlendFunc(src.factor_dest_rgb);
+        blend.a_equation = MaxwellToGL::BlendEquation(src.equation_a);
+        blend.src_a_func = MaxwellToGL::BlendFunc(src.factor_source_a);
+        blend.dst_a_func = MaxwellToGL::BlendFunc(src.factor_dest_a);
     }
 }
 
 void RasterizerOpenGL::SyncLogicOpState() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
 
-    // TODO(Subv): Support more than just render target 0.
     state.logic_op.enabled = regs.logic_op.enable != 0;
 
     if (!state.logic_op.enabled)
@@ -1108,19 +1121,21 @@ void RasterizerOpenGL::SyncLogicOpState() {
 }
 
 void RasterizerOpenGL::SyncScissorTest() {
-    // TODO: what is the correct behavior here, a single scissor for all targets
-    // or scissor disabled for the rest of the targets?
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-    state.scissor.enabled = (regs.scissor_test.enable != 0);
-    if (regs.scissor_test.enable == 0) {
-        return;
+    for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumViewports; i++) {
+        const auto& src = regs.scissor_test[i];
+        auto& dst = state.viewports[i].scissor;
+        dst.enabled = (src.enable != 0);
+        if (dst.enabled == 0) {
+            return;
+        }
+        const u32 width = src.max_x - src.min_x;
+        const u32 height = src.max_y - src.min_y;
+        dst.x = src.min_x;
+        dst.y = src.min_y;
+        dst.width = width;
+        dst.height = height;
     }
-    const u32 width = regs.scissor_test.max_x - regs.scissor_test.min_x;
-    const u32 height = regs.scissor_test.max_y - regs.scissor_test.min_y;
-    state.scissor.x = regs.scissor_test.min_x;
-    state.scissor.y = regs.scissor_test.min_y;
-    state.scissor.width = width;
-    state.scissor.height = height;
 }
 
 void RasterizerOpenGL::SyncTransformFeedback() {
@@ -1134,11 +1149,7 @@ void RasterizerOpenGL::SyncTransformFeedback() {
 
 void RasterizerOpenGL::SyncPointState() {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-
-    // TODO(Rodrigo): Most games do not set a point size. I think this is a case of a
-    // register carrying a default value. For now, if the point size is zero, assume it's
-    // OpenGL's default (1).
-    state.point.size = regs.point_size == 0 ? 1 : regs.point_size;
+    state.point.size = regs.point_size;
 }
 
 void RasterizerOpenGL::CheckAlphaTests() {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -88,7 +88,7 @@ private:
         /// SamplerInfo struct.
         void Create();
         /// Syncs the sampler object with the config, updating any necessary state.
-        void SyncWithConfig(const Tegra::Texture::FullTextureInfo& info);
+        void SyncWithConfig(const Tegra::Texture::TSCEntry& info);
 
     private:
         Tegra::Texture::TextureFilter mag_filter;
@@ -100,6 +100,10 @@ private:
         bool uses_depth_compare;
         Tegra::Texture::DepthCompareFunc depth_compare_func;
         GLvec4 border_color;
+        float min_lod;
+        float max_lod;
+        float lod_bias;
+        float max_anisotropic;
     };
 
     /**
@@ -160,6 +164,12 @@ private:
     /// Syncs the LogicOp state to match the guest state
     void SyncLogicOpState();
 
+    /// Syncs the the color clamp state
+    void SyncFragmentColorClampState();
+
+    /// Syncs the alpha coverage and alpha to one
+    void SyncMultiSampleState();
+
     /// Syncs the scissor test state to match the guest state
     void SyncScissorTest();
 

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -381,11 +381,8 @@ void MortonCopy(u32 stride, u32 block_height, u32 height, u32 block_depth, u32 d
     const u32 tile_size_y{GetDefaultBlockHeight(format)};
 
     if (morton_to_gl) {
-        const std::vector<u8> data =
-            Tegra::Texture::UnswizzleTexture(addr, tile_size_x, tile_size_y, bytes_per_pixel,
-                                             stride, height, depth, block_height, block_depth);
-        const std::size_t size_to_copy{std::min(gl_buffer_size, data.size())};
-        memcpy(gl_buffer, data.data(), size_to_copy);
+        Tegra::Texture::UnswizzleTexture(gl_buffer, addr, tile_size_x, tile_size_y, bytes_per_pixel,
+                                         stride, height, depth, block_height, block_depth);
     } else {
         Tegra::Texture::CopySwizzledData((stride + tile_size_x - 1) / tile_size_x,
                                          (height + tile_size_y - 1) / tile_size_y, depth,
@@ -1278,6 +1275,31 @@ Surface RasterizerCacheOpenGL::GetUncachedSurface(const SurfaceParams& params) {
     return surface;
 }
 
+void RasterizerCacheOpenGL::FastLayeredCopySurface(const Surface& src_surface,
+                                                   const Surface& dst_surface) {
+    const auto& init_params{src_surface->GetSurfaceParams()};
+    const auto& dst_params{dst_surface->GetSurfaceParams()};
+    VAddr address = init_params.addr;
+    const std::size_t layer_size = dst_params.LayerMemorySize();
+    for (u32 layer = 0; layer < dst_params.depth; layer++) {
+        for (u32 mipmap = 0; mipmap < dst_params.max_mip_level; mipmap++) {
+            const VAddr sub_address = address + dst_params.GetMipmapLevelOffset(mipmap);
+            const Surface& copy = TryGet(sub_address);
+            if (!copy)
+                continue;
+            const auto& src_params{copy->GetSurfaceParams()};
+            const u32 width{std::min(src_params.width, dst_params.MipWidth(mipmap))};
+            const u32 height{std::min(src_params.height, dst_params.MipHeight(mipmap))};
+
+            glCopyImageSubData(copy->Texture().handle, SurfaceTargetToGL(src_params.target), 0, 0,
+                               0, 0, dst_surface->Texture().handle,
+                               SurfaceTargetToGL(dst_params.target), mipmap, 0, 0, layer, width,
+                               height, 1);
+        }
+        address += layer_size;
+    }
+}
+
 void RasterizerCacheOpenGL::FermiCopySurface(
     const Tegra::Engines::Fermi2D::Regs::Surface& src_config,
     const Tegra::Engines::Fermi2D::Regs::Surface& dst_config) {
@@ -1343,11 +1365,13 @@ Surface RasterizerCacheOpenGL::RecreateSurface(const Surface& old_surface,
             CopySurface(old_surface, new_surface, copy_pbo.handle);
         }
         break;
-    case SurfaceTarget::TextureCubemap:
     case SurfaceTarget::Texture3D:
+        AccurateCopySurface(old_surface, new_surface);
+        break;
+    case SurfaceTarget::TextureCubemap:
     case SurfaceTarget::Texture2DArray:
     case SurfaceTarget::TextureCubeArray:
-        AccurateCopySurface(old_surface, new_surface);
+        FastLayeredCopySurface(old_surface, new_surface);
         break;
     default:
         LOG_CRITICAL(Render_OpenGL, "Unimplemented surface target={}",

src/video_core/renderer_opengl/gl_rasterizer_cache.h

@@ -350,6 +350,7 @@ private:
 
     /// Performs a slow but accurate surface copy, flushing to RAM and reinterpreting the data
     void AccurateCopySurface(const Surface& src_surface, const Surface& dst_surface);
+    void FastLayeredCopySurface(const Surface& src_surface, const Surface& dst_surface);
 
     /// The surface reserve is a "backup" cache, this is where we put unique surfaces that have
     /// previously been used. This is to prevent surfaces from being constantly created and

src/video_core/renderer_opengl/gl_shader_manager.h

@@ -67,6 +67,7 @@ public:
         glUseProgramStages(pipeline.handle, GL_FRAGMENT_SHADER_BIT, fs);
         state.draw.shader_program = 0;
         state.draw.program_pipeline = pipeline.handle;
+        state.geometry_shaders.enabled = (gs != 0);
     }
 
 private:

src/video_core/renderer_opengl/gl_state.cpp

@@ -14,7 +14,10 @@ OpenGLState OpenGLState::cur_state;
 bool OpenGLState::s_rgb_used;
 OpenGLState::OpenGLState() {
     // These all match default OpenGL values
+    geometry_shaders.enabled = false;
     framebuffer_srgb.enabled = false;
+    multisample_control.alpha_to_coverage = false;
+    multisample_control.alpha_to_one = false;
     cull.enabled = false;
     cull.mode = GL_BACK;
     cull.front_face = GL_CCW;
@@ -50,12 +53,12 @@ OpenGLState::OpenGLState() {
         item.height = 0;
         item.depth_range_near = 0.0f;
         item.depth_range_far = 1.0f;
+        item.scissor.enabled = false;
+        item.scissor.x = 0;
+        item.scissor.y = 0;
+        item.scissor.width = 0;
+        item.scissor.height = 0;
     }
-    scissor.enabled = false;
-    scissor.x = 0;
-    scissor.y = 0;
-    scissor.width = 0;
-    scissor.height = 0;
     for (auto& item : blend) {
         item.enabled = true;
         item.rgb_equation = GL_FUNC_ADD;
@@ -88,6 +91,7 @@ OpenGLState::OpenGLState() {
     clip_distance = {};
 
     point.size = 1;
+    fragment_color_clamp.enabled = false;
 }
 
 void OpenGLState::ApplyDefaultState() {
@@ -136,7 +140,7 @@ void OpenGLState::ApplyCulling() const {
 }
 
 void OpenGLState::ApplyColorMask() const {
-    if (GLAD_GL_ARB_viewport_array) {
+    if (GLAD_GL_ARB_viewport_array && independant_blend.enabled) {
         for (size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumRenderTargets; i++) {
             const auto& updated = color_mask[i];
             const auto& current = cur_state.color_mask[i];
@@ -230,26 +234,10 @@ void OpenGLState::ApplyStencilTest() const {
     }
 }
 
-void OpenGLState::ApplyScissor() const {
-    const bool scissor_changed = scissor.enabled != cur_state.scissor.enabled;
-    if (scissor_changed) {
-        if (scissor.enabled) {
-            glEnable(GL_SCISSOR_TEST);
-        } else {
-            glDisable(GL_SCISSOR_TEST);
-        }
-    }
-    if (scissor.enabled &&
-        (scissor_changed || scissor.x != cur_state.scissor.x || scissor.y != cur_state.scissor.y ||
-         scissor.width != cur_state.scissor.width || scissor.height != cur_state.scissor.height)) {
-        glScissor(scissor.x, scissor.y, scissor.width, scissor.height);
-    }
-}
-
 void OpenGLState::ApplyViewport() const {
-    if (GLAD_GL_ARB_viewport_array) {
-        for (GLuint i = 0;
-             i < static_cast<GLuint>(Tegra::Engines::Maxwell3D::Regs::NumRenderTargets); i++) {
+    if (GLAD_GL_ARB_viewport_array && geometry_shaders.enabled) {
+        for (GLuint i = 0; i < static_cast<GLuint>(Tegra::Engines::Maxwell3D::Regs::NumViewports);
+             i++) {
             const auto& current = cur_state.viewports[i];
             const auto& updated = viewports[i];
             if (updated.x != current.x || updated.y != current.y ||
@@ -260,6 +248,22 @@ void OpenGLState::ApplyViewport() const {
                 updated.depth_range_far != current.depth_range_far) {
                 glDepthRangeIndexed(i, updated.depth_range_near, updated.depth_range_far);
             }
+            const bool scissor_changed = updated.scissor.enabled != current.scissor.enabled;
+            if (scissor_changed) {
+                if (updated.scissor.enabled) {
+                    glEnablei(GL_SCISSOR_TEST, i);
+                } else {
+                    glDisablei(GL_SCISSOR_TEST, i);
+                }
+            }
+            if (updated.scissor.enabled &&
+                (scissor_changed || updated.scissor.x != current.scissor.x ||
+                 updated.scissor.y != current.scissor.y ||
+                 updated.scissor.width != current.scissor.width ||
+                 updated.scissor.height != current.scissor.height)) {
+                glScissorIndexed(i, updated.scissor.x, updated.scissor.y, updated.scissor.width,
+                                 updated.scissor.height);
+            }
         }
     } else {
         const auto& current = cur_state.viewports[0];
@@ -273,6 +277,21 @@ void OpenGLState::ApplyViewport() const {
             updated.depth_range_far != current.depth_range_far) {
             glDepthRange(updated.depth_range_near, updated.depth_range_far);
         }
+        const bool scissor_changed = updated.scissor.enabled != current.scissor.enabled;
+        if (scissor_changed) {
+            if (updated.scissor.enabled) {
+                glEnable(GL_SCISSOR_TEST);
+            } else {
+                glDisable(GL_SCISSOR_TEST);
+            }
+        }
+        if (updated.scissor.enabled && (scissor_changed || updated.scissor.x != current.scissor.x ||
+                                        updated.scissor.y != current.scissor.y ||
+                                        updated.scissor.width != current.scissor.width ||
+                                        updated.scissor.height != current.scissor.height)) {
+            glScissor(updated.scissor.x, updated.scissor.y, updated.scissor.width,
+                      updated.scissor.height);
+        }
     }
 }
 
@@ -290,27 +309,16 @@ void OpenGLState::ApplyGlobalBlending() const {
     if (!updated.enabled) {
         return;
     }
-    if (updated.separate_alpha) {
-        if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
-            updated.dst_rgb_func != current.dst_rgb_func ||
-            updated.src_a_func != current.src_a_func || updated.dst_a_func != current.dst_a_func) {
-            glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
-                                updated.dst_a_func);
-        }
+    if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
+        updated.dst_rgb_func != current.dst_rgb_func || updated.src_a_func != current.src_a_func ||
+        updated.dst_a_func != current.dst_a_func) {
+        glBlendFuncSeparate(updated.src_rgb_func, updated.dst_rgb_func, updated.src_a_func,
+                            updated.dst_a_func);
+    }
 
-        if (blend_changed || updated.rgb_equation != current.rgb_equation ||
-            updated.a_equation != current.a_equation) {
-            glBlendEquationSeparate(updated.rgb_equation, updated.a_equation);
-        }
-    } else {
-        if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
-            updated.dst_rgb_func != current.dst_rgb_func) {
-            glBlendFunc(updated.src_rgb_func, updated.dst_rgb_func);
-        }
-
-        if (blend_changed || updated.rgb_equation != current.rgb_equation) {
-            glBlendEquation(updated.rgb_equation);
-        }
+    if (blend_changed || updated.rgb_equation != current.rgb_equation ||
+        updated.a_equation != current.a_equation) {
+        glBlendEquationSeparate(updated.rgb_equation, updated.a_equation);
     }
 }
 
@@ -328,29 +336,17 @@ void OpenGLState::ApplyTargetBlending(std::size_t target, bool force) const {
     if (!updated.enabled) {
         return;
     }
-    if (updated.separate_alpha) {
-        if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
-            updated.dst_rgb_func != current.dst_rgb_func ||
-            updated.src_a_func != current.src_a_func || updated.dst_a_func != current.dst_a_func) {
-            glBlendFuncSeparateiARB(static_cast<GLuint>(target), updated.src_rgb_func,
-                                    updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
-        }
+    if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
+        updated.dst_rgb_func != current.dst_rgb_func || updated.src_a_func != current.src_a_func ||
+        updated.dst_a_func != current.dst_a_func) {
+        glBlendFuncSeparateiARB(static_cast<GLuint>(target), updated.src_rgb_func,
+                                updated.dst_rgb_func, updated.src_a_func, updated.dst_a_func);
+    }
 
-        if (blend_changed || updated.rgb_equation != current.rgb_equation ||
-            updated.a_equation != current.a_equation) {
-            glBlendEquationSeparateiARB(static_cast<GLuint>(target), updated.rgb_equation,
-                                        updated.a_equation);
-        }
-    } else {
-        if (blend_changed || updated.src_rgb_func != current.src_rgb_func ||
-            updated.dst_rgb_func != current.dst_rgb_func) {
-            glBlendFunciARB(static_cast<GLuint>(target), updated.src_rgb_func,
-                            updated.dst_rgb_func);
-        }
-
-        if (blend_changed || updated.rgb_equation != current.rgb_equation) {
-            glBlendEquationiARB(static_cast<GLuint>(target), updated.rgb_equation);
-        }
+    if (blend_changed || updated.rgb_equation != current.rgb_equation ||
+        updated.a_equation != current.a_equation) {
+        glBlendEquationSeparateiARB(static_cast<GLuint>(target), updated.rgb_equation,
+                                    updated.a_equation);
     }
 }
 
@@ -481,9 +477,29 @@ void OpenGLState::Apply() const {
     if (point.size != cur_state.point.size) {
         glPointSize(point.size);
     }
+    if (GLAD_GL_ARB_color_buffer_float) {
+        if (fragment_color_clamp.enabled != cur_state.fragment_color_clamp.enabled) {
+            glClampColor(GL_CLAMP_FRAGMENT_COLOR_ARB,
+                         fragment_color_clamp.enabled ? GL_TRUE : GL_FALSE);
+        }
+    }
+    if (multisample_control.alpha_to_coverage != cur_state.multisample_control.alpha_to_coverage) {
+        if (multisample_control.alpha_to_coverage) {
+            glEnable(GL_SAMPLE_ALPHA_TO_COVERAGE);
+        } else {
+            glDisable(GL_SAMPLE_ALPHA_TO_COVERAGE);
+        }
+    }
+    if (multisample_control.alpha_to_one != cur_state.multisample_control.alpha_to_one) {
+        if (multisample_control.alpha_to_one) {
+            glEnable(GL_SAMPLE_ALPHA_TO_ONE);
+        } else {
+            glDisable(GL_SAMPLE_ALPHA_TO_ONE);
+        }
+    }
+
     ApplyColorMask();
     ApplyViewport();
-    ApplyScissor();
     ApplyStencilTest();
     ApplySRgb();
     ApplyCulling();

src/video_core/renderer_opengl/gl_state.h

@@ -39,6 +39,19 @@ public:
         bool enabled; // GL_FRAMEBUFFER_SRGB
     } framebuffer_srgb;
 
+    struct {
+        bool alpha_to_coverage; // GL_ALPHA_TO_COVERAGE
+        bool alpha_to_one;      // GL_ALPHA_TO_ONE
+    } multisample_control;
+
+    struct {
+        bool enabled; // GL_CLAMP_FRAGMENT_COLOR_ARB
+    } fragment_color_clamp;
+
+    struct {
+        bool enabled; // viewports arrays are only supported when geometry shaders are enabled.
+    } geometry_shaders;
+
     struct {
         bool enabled;      // GL_CULL_FACE
         GLenum mode;       // GL_CULL_FACE_MODE
@@ -79,7 +92,6 @@ public:
 
     struct Blend {
         bool enabled;        // GL_BLEND
-        bool separate_alpha; // Independent blend enabled
         GLenum rgb_equation; // GL_BLEND_EQUATION_RGB
         GLenum a_equation;   // GL_BLEND_EQUATION_ALPHA
         GLenum src_rgb_func; // GL_BLEND_SRC_RGB
@@ -150,16 +162,15 @@ public:
         GLfloat height;
         GLfloat depth_range_near; // GL_DEPTH_RANGE
         GLfloat depth_range_far;  // GL_DEPTH_RANGE
+        struct {
+            bool enabled; // GL_SCISSOR_TEST
+            GLint x;
+            GLint y;
+            GLsizei width;
+            GLsizei height;
+        } scissor;
     };
-    std::array<viewport, Tegra::Engines::Maxwell3D::Regs::NumRenderTargets> viewports;
-
-    struct {
-        bool enabled; // GL_SCISSOR_TEST
-        GLint x;
-        GLint y;
-        GLsizei width;
-        GLsizei height;
-    } scissor;
+    std::array<viewport, Tegra::Engines::Maxwell3D::Regs::NumViewports> viewports;
 
     struct {
         float size; // GL_POINT_SIZE
@@ -214,7 +225,6 @@ private:
     void ApplyLogicOp() const;
     void ApplyTextures() const;
     void ApplySamplers() const;
-    void ApplyScissor() const;
 };
 
 } // namespace OpenGL

src/video_core/renderer_opengl/maxwell_to_gl.h

@@ -180,6 +180,12 @@ inline GLenum WrapMode(Tegra::Texture::WrapMode wrap_mode) {
         return GL_CLAMP_TO_BORDER;
     case Tegra::Texture::WrapMode::MirrorOnceClampToEdge:
         return GL_MIRROR_CLAMP_TO_EDGE;
+    case Tegra::Texture::WrapMode::MirrorOnceBorder:
+        if (GL_EXT_texture_mirror_clamp) {
+            return GL_MIRROR_CLAMP_TO_BORDER_EXT;
+        } else {
+            return GL_MIRROR_CLAMP_TO_EDGE;
+        }
     }
     LOG_ERROR(Render_OpenGL, "Unimplemented texture wrap mode={}", static_cast<u32>(wrap_mode));
     return GL_REPEAT;

src/video_core/textures/decoders.cpp

@@ -37,8 +37,14 @@ struct alignas(64) SwizzleTable {
     std::array<std::array<u16, M>, N> values{};
 };
 
-constexpr auto legacy_swizzle_table = SwizzleTable<8, 64, 1>();
-constexpr auto fast_swizzle_table = SwizzleTable<8, 4, 16>();
+constexpr u32 gob_size_x = 64;
+constexpr u32 gob_size_y = 8;
+constexpr u32 gob_size_z = 1;
+constexpr u32 gob_size = gob_size_x * gob_size_y * gob_size_z;
+constexpr u32 fast_swizzle_align = 16;
+
+constexpr auto legacy_swizzle_table = SwizzleTable<gob_size_y, gob_size_x, gob_size_z>();
+constexpr auto fast_swizzle_table = SwizzleTable<gob_size_y, 4, fast_swizzle_align>();
 
 /**
  * This function manages ALL the GOBs(Group of Bytes) Inside a single block.
@@ -52,10 +58,7 @@ void PreciseProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, con
                          const u32 bytes_per_pixel, const u32 out_bytes_per_pixel) {
     std::array<u8*, 2> data_ptrs;
     u32 z_address = tile_offset;
-    const u32 gob_size_x = 64;
-    const u32 gob_size_y = 8;
-    const u32 gob_size_z = 1;
-    const u32 gob_size = gob_size_x * gob_size_y * gob_size_z;
+
     for (u32 z = z_start; z < z_end; z++) {
         u32 y_address = z_address;
         u32 pixel_base = layer_z * z + y_start * stride_x;
@@ -90,23 +93,19 @@ void FastProcessBlock(u8* const swizzled_data, u8* const unswizzled_data, const
     u32 z_address = tile_offset;
     const u32 x_startb = x_start * bytes_per_pixel;
     const u32 x_endb = x_end * bytes_per_pixel;
-    constexpr u32 copy_size = 16;
-    constexpr u32 gob_size_x = 64;
-    constexpr u32 gob_size_y = 8;
-    constexpr u32 gob_size_z = 1;
-    const u32 gob_size = gob_size_x * gob_size_y * gob_size_z;
+
     for (u32 z = z_start; z < z_end; z++) {
         u32 y_address = z_address;
         u32 pixel_base = layer_z * z + y_start * stride_x;
         for (u32 y = y_start; y < y_end; y++) {
             const auto& table = fast_swizzle_table[y % gob_size_y];
-            for (u32 xb = x_startb; xb < x_endb; xb += copy_size) {
-                const u32 swizzle_offset{y_address + table[(xb / copy_size) % 4]};
+            for (u32 xb = x_startb; xb < x_endb; xb += fast_swizzle_align) {
+                const u32 swizzle_offset{y_address + table[(xb / fast_swizzle_align) % 4]};
                 const u32 out_x = xb * out_bytes_per_pixel / bytes_per_pixel;
                 const u32 pixel_index{out_x + pixel_base};
                 data_ptrs[unswizzle] = swizzled_data + swizzle_offset;
                 data_ptrs[!unswizzle] = unswizzled_data + pixel_index;
-                std::memcpy(data_ptrs[0], data_ptrs[1], copy_size);
+                std::memcpy(data_ptrs[0], data_ptrs[1], fast_swizzle_align);
             }
             pixel_base += stride_x;
             if ((y + 1) % gob_size_y == 0)
@@ -132,17 +131,15 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool
     auto div_ceil = [](const u32 x, const u32 y) { return ((x + y - 1) / y); };
     const u32 stride_x = width * out_bytes_per_pixel;
     const u32 layer_z = height * stride_x;
-    constexpr u32 gob_x_bytes = 64;
-    const u32 gob_elements_x = gob_x_bytes / bytes_per_pixel;
-    constexpr u32 gob_elements_y = 8;
-    constexpr u32 gob_elements_z = 1;
+    const u32 gob_elements_x = gob_size_x / bytes_per_pixel;
+    constexpr u32 gob_elements_y = gob_size_y;
+    constexpr u32 gob_elements_z = gob_size_z;
     const u32 block_x_elements = gob_elements_x;
     const u32 block_y_elements = gob_elements_y * block_height;
     const u32 block_z_elements = gob_elements_z * block_depth;
     const u32 blocks_on_x = div_ceil(width, block_x_elements);
     const u32 blocks_on_y = div_ceil(height, block_y_elements);
     const u32 blocks_on_z = div_ceil(depth, block_z_elements);
-    constexpr u32 gob_size = gob_x_bytes * gob_elements_y * gob_elements_z;
     const u32 xy_block_size = gob_size * block_height;
     const u32 block_size = xy_block_size * block_depth;
     u32 tile_offset = 0;
@@ -173,7 +170,7 @@ void SwizzledData(u8* const swizzled_data, u8* const unswizzled_data, const bool
 void CopySwizzledData(u32 width, u32 height, u32 depth, u32 bytes_per_pixel,
                       u32 out_bytes_per_pixel, u8* const swizzled_data, u8* const unswizzled_data,
                       bool unswizzle, u32 block_height, u32 block_depth) {
-    if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % 16 == 0) {
+    if (bytes_per_pixel % 3 != 0 && (width * bytes_per_pixel) % fast_swizzle_align == 0) {
         SwizzledData<true>(swizzled_data, unswizzled_data, unswizzle, width, height, depth,
                            bytes_per_pixel, out_bytes_per_pixel, block_height, block_depth);
     } else {
@@ -229,29 +226,38 @@ u32 BytesPerPixel(TextureFormat format) {
     }
 }
 
+void UnswizzleTexture(u8* const unswizzled_data, VAddr address, u32 tile_size_x, u32 tile_size_y,
+                      u32 bytes_per_pixel, u32 width, u32 height, u32 depth, u32 block_height,
+                      u32 block_depth) {
+    CopySwizzledData((width + tile_size_x - 1) / tile_size_x,
+                     (height + tile_size_y - 1) / tile_size_y, depth, bytes_per_pixel,
+                     bytes_per_pixel, Memory::GetPointer(address), unswizzled_data, true,
+                     block_height, block_depth);
+}
+
 std::vector<u8> UnswizzleTexture(VAddr address, u32 tile_size_x, u32 tile_size_y,
                                  u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                                  u32 block_height, u32 block_depth) {
     std::vector<u8> unswizzled_data(width * height * depth * bytes_per_pixel);
-    CopySwizzledData((width + tile_size_x - 1) / tile_size_x,
-                     (height + tile_size_y - 1) / tile_size_y, depth, bytes_per_pixel,
-                     bytes_per_pixel, Memory::GetPointer(address), unswizzled_data.data(), true,
-                     block_height, block_depth);
+    UnswizzleTexture(unswizzled_data.data(), address, tile_size_x, tile_size_y, bytes_per_pixel,
+                     width, height, depth, block_height, block_depth);
     return unswizzled_data;
 }
 
 void SwizzleSubrect(u32 subrect_width, u32 subrect_height, u32 source_pitch, u32 swizzled_width,
                     u32 bytes_per_pixel, VAddr swizzled_data, VAddr unswizzled_data,
                     u32 block_height) {
-    const u32 image_width_in_gobs{(swizzled_width * bytes_per_pixel + 63) / 64};
+    const u32 image_width_in_gobs{(swizzled_width * bytes_per_pixel + (gob_size_x - 1)) /
+                                  gob_size_x};
     for (u32 line = 0; line < subrect_height; ++line) {
         const u32 gob_address_y =
-            (line / (8 * block_height)) * 512 * block_height * image_width_in_gobs +
-            (line % (8 * block_height) / 8) * 512;
-        const auto& table = legacy_swizzle_table[line % 8];
+            (line / (gob_size_y * block_height)) * gob_size * block_height * image_width_in_gobs +
+            ((line % (gob_size_y * block_height)) / gob_size_y) * gob_size;
+        const auto& table = legacy_swizzle_table[line % gob_size_y];
         for (u32 x = 0; x < subrect_width; ++x) {
-            const u32 gob_address = gob_address_y + (x * bytes_per_pixel / 64) * 512 * block_height;
-            const u32 swizzled_offset = gob_address + table[(x * bytes_per_pixel) % 64];
+            const u32 gob_address =
+                gob_address_y + (x * bytes_per_pixel / gob_size_x) * gob_size * block_height;
+            const u32 swizzled_offset = gob_address + table[(x * bytes_per_pixel) % gob_size_x];
             const VAddr source_line = unswizzled_data + line * source_pitch + x * bytes_per_pixel;
             const VAddr dest_addr = swizzled_data + swizzled_offset;
 
@@ -265,13 +271,13 @@ void UnswizzleSubrect(u32 subrect_width, u32 subrect_height, u32 dest_pitch, u32
                       u32 block_height, u32 offset_x, u32 offset_y) {
     for (u32 line = 0; line < subrect_height; ++line) {
         const u32 y2 = line + offset_y;
-        const u32 gob_address_y =
-            (y2 / (8 * block_height)) * 512 * block_height + (y2 % (8 * block_height) / 8) * 512;
-        const auto& table = legacy_swizzle_table[y2 % 8];
+        const u32 gob_address_y = (y2 / (gob_size_y * block_height)) * gob_size * block_height +
+                                  ((y2 % (gob_size_y * block_height)) / gob_size_y) * gob_size;
+        const auto& table = legacy_swizzle_table[y2 % gob_size_y];
         for (u32 x = 0; x < subrect_width; ++x) {
             const u32 x2 = (x + offset_x) * bytes_per_pixel;
-            const u32 gob_address = gob_address_y + (x2 / 64) * 512 * block_height;
-            const u32 swizzled_offset = gob_address + table[x2 % 64];
+            const u32 gob_address = gob_address_y + (x2 / gob_size_x) * gob_size * block_height;
+            const u32 swizzled_offset = gob_address + table[x2 % gob_size_x];
             const VAddr dest_line = unswizzled_data + line * dest_pitch + x * bytes_per_pixel;
             const VAddr source_addr = swizzled_data + swizzled_offset;
 
@@ -325,12 +331,9 @@ std::vector<u8> DecodeTexture(const std::vector<u8>& texture_data, TextureFormat
 std::size_t CalculateSize(bool tiled, u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
                           u32 block_height, u32 block_depth) {
     if (tiled) {
-        constexpr u32 gobs_in_x = 64;
-        constexpr u32 gobs_in_y = 8;
-        constexpr u32 gobs_in_z = 1;
-        const u32 aligned_width = Common::AlignUp(width * bytes_per_pixel, gobs_in_x);
-        const u32 aligned_height = Common::AlignUp(height, gobs_in_y * block_height);
-        const u32 aligned_depth = Common::AlignUp(depth, gobs_in_z * block_depth);
+        const u32 aligned_width = Common::AlignUp(width * bytes_per_pixel, gob_size_x);
+        const u32 aligned_height = Common::AlignUp(height, gob_size_y * block_height);
+        const u32 aligned_depth = Common::AlignUp(depth, gob_size_z * block_depth);
         return aligned_width * aligned_height * aligned_depth;
     } else {
         return width * height * depth * bytes_per_pixel;

src/video_core/textures/decoders.h

@@ -16,6 +16,13 @@ inline std::size_t GetGOBSize() {
     return 512;
 }
 
+/**
+ * Unswizzles a swizzled texture without changing its format.
+ */
+void UnswizzleTexture(u8* unswizzled_data, VAddr address, u32 tile_size_x, u32 tile_size_y,
+                      u32 bytes_per_pixel, u32 width, u32 height, u32 depth,
+                      u32 block_height = TICEntry::DefaultBlockHeight,
+                      u32 block_depth = TICEntry::DefaultBlockHeight);
 /**
  * Unswizzles a swizzled texture without changing its format.
  */

src/video_core/textures/texture.h

@@ -190,6 +190,7 @@ struct TICEntry {
     union {
         BitField<0, 4, u32> res_min_mip_level;
         BitField<4, 4, u32> res_max_mip_level;
+        BitField<12, 12, u32> min_lod_clamp;
     };
 
     GPUVAddr Address() const {
@@ -284,13 +285,25 @@ struct TSCEntry {
         BitField<6, 3, WrapMode> wrap_p;
         BitField<9, 1, u32> depth_compare_enabled;
         BitField<10, 3, DepthCompareFunc> depth_compare_func;
+        BitField<13, 1, u32> srgb_conversion;
+        BitField<20, 3, u32> max_anisotropy;
     };
     union {
         BitField<0, 2, TextureFilter> mag_filter;
         BitField<4, 2, TextureFilter> min_filter;
         BitField<6, 2, TextureMipmapFilter> mip_filter;
+        BitField<9, 1, u32> cubemap_interface_filtering;
+        BitField<12, 13, u32> mip_lod_bias;
+    };
+    union {
+        BitField<0, 12, u32> min_lod_clamp;
+        BitField<12, 12, u32> max_lod_clamp;
+        BitField<24, 8, u32> srgb_border_color_r;
+    };
+    union {
+        BitField<12, 8, u32> srgb_border_color_g;
+        BitField<20, 8, u32> srgb_border_color_b;
     };
-    INSERT_PADDING_BYTES(8);
     float border_color_r;
     float border_color_g;
     float border_color_b;

src/yuzu/configuration/configure_gamelist.cpp

@@ -36,6 +36,16 @@ ConfigureGameList::ConfigureGameList(QWidget* parent)
     InitializeRowComboBoxes();
 
     this->setConfiguration();
+
+    // Force game list reload if any of the relevant settings are changed.
+    connect(ui->show_unknown, &QCheckBox::stateChanged, this,
+            &ConfigureGameList::RequestGameListUpdate);
+    connect(ui->icon_size_combobox, QOverload<int>::of(&QComboBox::currentIndexChanged), this,
+            &ConfigureGameList::RequestGameListUpdate);
+    connect(ui->row_1_text_combobox, QOverload<int>::of(&QComboBox::currentIndexChanged), this,
+            &ConfigureGameList::RequestGameListUpdate);
+    connect(ui->row_2_text_combobox, QOverload<int>::of(&QComboBox::currentIndexChanged), this,
+            &ConfigureGameList::RequestGameListUpdate);
 }
 
 ConfigureGameList::~ConfigureGameList() = default;
@@ -49,6 +59,10 @@ void ConfigureGameList::applyConfiguration() {
     Settings::Apply();
 }
 
+void ConfigureGameList::RequestGameListUpdate() {
+    UISettings::values.is_game_list_reload_pending.exchange(true);
+}
+
 void ConfigureGameList::setConfiguration() {
     ui->show_unknown->setChecked(UISettings::values.show_unknown);
     ui->show_add_ons->setChecked(UISettings::values.show_add_ons);

src/yuzu/configuration/configure_gamelist.h

@@ -21,6 +21,8 @@ public:
     void applyConfiguration();
 
 private:
+    void RequestGameListUpdate();
+
     void setConfiguration();
 
     void changeEvent(QEvent*) override;

src/yuzu/configuration/configure_general.cpp

@@ -23,6 +23,9 @@ ConfigureGeneral::ConfigureGeneral(QWidget* parent)
 
     this->setConfiguration();
 
+    connect(ui->toggle_deepscan, &QCheckBox::stateChanged, this,
+            [] { UISettings::values.is_game_list_reload_pending.exchange(true); });
+
     ui->use_cpu_jit->setEnabled(!Core::System::GetInstance().IsPoweredOn());
 }
 

src/yuzu/main.cpp

@@ -1341,7 +1341,13 @@ void GMainWindow::OnConfigure() {
             UpdateUITheme();
         if (UISettings::values.enable_discord_presence != old_discord_presence)
             SetDiscordEnabled(UISettings::values.enable_discord_presence);
-        game_list->PopulateAsync(UISettings::values.gamedir, UISettings::values.gamedir_deepscan);
+
+        const auto reload = UISettings::values.is_game_list_reload_pending.exchange(false);
+        if (reload) {
+            game_list->PopulateAsync(UISettings::values.gamedir,
+                                     UISettings::values.gamedir_deepscan);
+        }
+
         config->Save();
     }
 }

src/yuzu/ui_settings.h

@@ -5,6 +5,7 @@
 #pragma once
 
 #include <array>
+#include <atomic>
 #include <vector>
 #include <QByteArray>
 #include <QString>
@@ -63,6 +64,7 @@ struct Values {
     uint32_t icon_size;
     uint8_t row_1_text_id;
     uint8_t row_2_text_id;
+    std::atomic_bool is_game_list_reload_pending{false};
 };
 
 extern Values values;

src/yuzu_cmd/config.cpp

@@ -139,6 +139,8 @@ void Config::ReadValues() {
         Settings::values.rng_seed = std::nullopt;
     }
 
+    Settings::values.language_index = sdl2_config->GetInteger("System", "language_index", 1);
+
     // Miscellaneous
     Settings::values.log_filter = sdl2_config->Get("Miscellaneous", "log_filter", "*:Trace");
     Settings::values.use_dev_keys = sdl2_config->GetBoolean("Miscellaneous", "use_dev_keys", false);