gl_device: Force GLASM on NVIDIA drivers 495-496

GLSL shaders currently do not render correctly on the recent NVIDIA
drivers. This adds a check that forces assembly shaders for these
drivers since they seem unaffected and adds a warning informing of the
decision.

Developers can disable the check by enabling graphics debugging.

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -263,7 +263,7 @@ void ARM_Dynarmic_64::Run() {
 }
 
 void ARM_Dynarmic_64::Step() {
-    cb->InterpreterFallback(jit->GetPC(), 1);
+    jit->Step();
 }
 
 ARM_Dynarmic_64::ARM_Dynarmic_64(System& system_, CPUInterrupts& interrupt_handlers_,

src/core/hle/kernel/k_scheduler.h

@@ -49,6 +49,11 @@ public:
     /// Gets the current running thread
     [[nodiscard]] KThread* GetCurrentThread() const;
 
+    /// Gets the idle thread
+    [[nodiscard]] KThread* GetIdleThread() const {
+        return idle_thread;
+    }
+
     /// Returns true if the scheduler is idle
     [[nodiscard]] bool IsIdle() const {
         return GetCurrentThread() == idle_thread;

src/core/hle/kernel/svc.cpp

@@ -886,7 +886,24 @@ static ResultCode GetInfo(Core::System& system, u64* result, u64 info_id, Handle
         *result = out_ticks;
         return ResultSuccess;
     }
+    case GetInfoType::IdleTickCount: {
+        if (handle == 0) {
+            LOG_ERROR(Kernel_SVC, "Thread handle does not exist, handle=0x{:08X}",
+                      static_cast<Handle>(handle));
+            return ResultInvalidHandle;
+        }
 
+        if (info_sub_id != 0xFFFFFFFFFFFFFFFF && info_sub_id != system.CurrentCoreIndex()) {
+            LOG_ERROR(Kernel_SVC, "Core is not the current core, got {}", info_sub_id);
+            return ResultInvalidCombination;
+        }
+
+        const auto& scheduler = *system.Kernel().CurrentScheduler();
+        const auto* const idle_thread = scheduler.GetIdleThread();
+
+        *result = idle_thread->GetCpuTime();
+        return ResultSuccess;
+    }
     default:
         LOG_ERROR(Kernel_SVC, "Unimplemented svcGetInfo id=0x{:016X}", info_id);
         return ResultInvalidEnumValue;

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

@@ -69,8 +69,7 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
     std::vector<Reloc> relocs(params.relocation_count);
     std::vector<u32> reloc_shifts(params.relocation_count);
     std::vector<SyncptIncr> syncpt_increments(params.syncpoint_count);
-    std::vector<SyncptIncr> wait_checks(params.syncpoint_count);
-    std::vector<Fence> fences(params.fence_count);
+    std::vector<u32> fence_thresholds(params.fence_count);
 
     // Slice input into their respective buffers
     std::size_t offset = sizeof(IoctlSubmit);
@@ -78,15 +77,13 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
     offset += SliceVectors(input, relocs, params.relocation_count, offset);
     offset += SliceVectors(input, reloc_shifts, params.relocation_count, offset);
     offset += SliceVectors(input, syncpt_increments, params.syncpoint_count, offset);
-    offset += SliceVectors(input, wait_checks, params.syncpoint_count, offset);
-    offset += SliceVectors(input, fences, params.fence_count, offset);
+    offset += SliceVectors(input, fence_thresholds, params.fence_count, offset);
 
     auto& gpu = system.GPU();
     if (gpu.UseNvdec()) {
         for (std::size_t i = 0; i < syncpt_increments.size(); i++) {
             const SyncptIncr& syncpt_incr = syncpt_increments[i];
-            fences[i].id = syncpt_incr.id;
-            fences[i].value =
+            fence_thresholds[i] =
                 syncpoint_manager.IncreaseSyncpoint(syncpt_incr.id, syncpt_incr.increments);
         }
     }
@@ -98,11 +95,6 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
                                   cmdlist.size() * sizeof(u32));
         gpu.PushCommandBuffer(cmdlist);
     }
-    if (gpu.UseNvdec()) {
-        fences[0].value = syncpoint_manager.IncreaseSyncpoint(fences[0].id, 1);
-        Tegra::ChCommandHeaderList cmdlist{{(4 << 28) | fences[0].id}};
-        gpu.PushCommandBuffer(cmdlist);
-    }
     std::memcpy(output.data(), &params, sizeof(IoctlSubmit));
     // Some games expect command_buffers to be written back
     offset = sizeof(IoctlSubmit);
@@ -110,8 +102,7 @@ NvResult nvhost_nvdec_common::Submit(const std::vector<u8>& input, std::vector<u
     offset += WriteVectors(output, relocs, offset);
     offset += WriteVectors(output, reloc_shifts, offset);
     offset += WriteVectors(output, syncpt_increments, offset);
-    offset += WriteVectors(output, wait_checks, offset);
-    offset += WriteVectors(output, fences, offset);
+    offset += WriteVectors(output, fence_thresholds, offset);
 
     return NvResult::Success;
 }

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

@@ -56,19 +56,16 @@ protected:
         s32 target{};
         s32 target_offset{};
     };
-    static_assert(sizeof(Reloc) == 0x10, "CommandBuffer has incorrect size");
+    static_assert(sizeof(Reloc) == 0x10, "Reloc has incorrect size");
 
     struct SyncptIncr {
         u32 id{};
         u32 increments{};
+        u32 unk0{};
+        u32 unk1{};
+        u32 unk2{};
     };
-    static_assert(sizeof(SyncptIncr) == 0x8, "CommandBuffer has incorrect size");
-
-    struct Fence {
-        u32 id{};
-        u32 value{};
-    };
-    static_assert(sizeof(Fence) == 0x8, "CommandBuffer has incorrect size");
+    static_assert(sizeof(SyncptIncr) == 0x14, "SyncptIncr has incorrect size");
 
     struct IoctlGetSyncpoint {
         // Input

src/video_core/renderer_opengl/gl_device.cpp

@@ -181,6 +181,21 @@ Device::Device() {
         LOG_ERROR(Render_OpenGL, "Assembly shaders enabled but not supported");
         shader_backend = Settings::ShaderBackend::GLSL;
     }
+
+    if (shader_backend == Settings::ShaderBackend::GLSL && is_nvidia &&
+        !Settings::values.renderer_debug) {
+        const std::string_view driver_version = version.substr(13);
+        const int version_major =
+            std::atoi(driver_version.substr(0, driver_version.find(".")).data());
+
+        if (version_major >= 495) {
+            LOG_WARNING(Render_OpenGL, "NVIDIA drivers 495 and later causes significant problems "
+                                       "with yuzu. Forcing GLASM as a mitigation.");
+            shader_backend = Settings::ShaderBackend::GLASM;
+            use_assembly_shaders = true;
+        }
+    }
+
     // Blocks AMD and Intel OpenGL drivers on Windows from using asynchronous shader compilation.
     use_asynchronous_shaders = Settings::values.use_asynchronous_shaders.GetValue() &&
                                !(is_amd || (is_intel && !is_linux));

src/video_core/renderer_vulkan/vk_state_tracker.cpp

@@ -54,6 +54,7 @@ void SetupDirtyViewports(Tables& tables) {
     FillBlock(tables[0], OFF(viewport_transform), NUM(viewport_transform), Viewports);
     FillBlock(tables[0], OFF(viewports), NUM(viewports), Viewports);
     tables[0][OFF(viewport_transform_enabled)] = Viewports;
+    tables[1][OFF(screen_y_control)] = Viewports;
 }
 
 void SetupDirtyScissors(Tables& tables) {

src/yuzu/configuration/configure_dialog.cpp

@@ -81,8 +81,11 @@ ConfigureDialog::ConfigureDialog(QWidget* parent, HotkeyRegistry& registry,
     SetConfiguration();
     PopulateSelectionList();
 
-    connect(ui->tabWidget, &QTabWidget::currentChanged, this,
-            [this]() { debug_tab_tab->SetCurrentIndex(0); });
+    connect(ui->tabWidget, &QTabWidget::currentChanged, this, [this](int index) {
+        if (index != -1) {
+            debug_tab_tab->SetCurrentIndex(0);
+        }
+    });
     connect(ui_tab.get(), &ConfigureUi::LanguageChanged, this, &ConfigureDialog::OnLanguageChanged);
     connect(ui->selectorList, &QListWidget::itemSelectionChanged, this,
             &ConfigureDialog::UpdateVisibleTabs);