kernel: avoid usage of bit_cast

Remove fake vertex bindings when dynamic state is enabled

src/audio_core/sink/cubeb_sink.cpp

@@ -302,11 +302,21 @@ std::vector<std::string> ListCubebSinkDevices(bool capture) {
     std::vector<std::string> device_list;
     cubeb* ctx;
 
+#ifdef _WIN32
+    auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
+#endif
+
     if (cubeb_init(&ctx, "yuzu Device Enumerator", nullptr) != CUBEB_OK) {
         LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
         return {};
     }
 
+#ifdef _WIN32
+    if (SUCCEEDED(com_init_result)) {
+        CoUninitialize();
+    }
+#endif
+
     auto type{capture ? CUBEB_DEVICE_TYPE_INPUT : CUBEB_DEVICE_TYPE_OUTPUT};
     cubeb_device_collection collection;
     if (cubeb_enumerate_devices(ctx, type, &collection) != CUBEB_OK) {
@@ -329,12 +339,22 @@ std::vector<std::string> ListCubebSinkDevices(bool capture) {
 u32 GetCubebLatency() {
     cubeb* ctx;
 
+#ifdef _WIN32
+    auto com_init_result = CoInitializeEx(nullptr, COINIT_MULTITHREADED);
+#endif
+
     if (cubeb_init(&ctx, "yuzu Latency Getter", nullptr) != CUBEB_OK) {
         LOG_CRITICAL(Audio_Sink, "cubeb_init failed");
         // Return a large latency so we choose SDL instead.
         return 10000u;
     }
 
+#ifdef _WIN32
+    if (SUCCEEDED(com_init_result)) {
+        CoUninitialize();
+    }
+#endif
+
     cubeb_stream_params params{};
     params.rate = TargetSampleRate;
     params.channels = 2;

src/core/hle/kernel/k_capabilities.cpp

@@ -203,23 +203,23 @@ Result KCapabilities::ProcessMapRegionCapability(const u32 cap, F f) {
 
 Result KCapabilities::MapRegion_(const u32 cap, KPageTable* page_table) {
     // Map each region into the process's page table.
-    R_RETURN(ProcessMapRegionCapability(
+    return ProcessMapRegionCapability(
         cap, [](KMemoryRegionType region_type, KMemoryPermission perm) -> Result {
             // R_RETURN(page_table->MapRegion(region_type, perm));
             UNIMPLEMENTED();
             R_SUCCEED();
-        }));
+        });
 }
 
 Result KCapabilities::CheckMapRegion(KernelCore& kernel, const u32 cap) {
     // Check that each region has a physical backing store.
-    R_RETURN(ProcessMapRegionCapability(
+    return ProcessMapRegionCapability(
         cap, [&](KMemoryRegionType region_type, KMemoryPermission perm) -> Result {
             R_UNLESS(kernel.MemoryLayout().GetPhysicalMemoryRegionTree().FindFirstDerived(
                          region_type) != nullptr,
                      ResultOutOfRange);
             R_SUCCEED();
-        }));
+        });
 }
 
 Result KCapabilities::SetInterruptPairCapability(const u32 cap) {

src/core/hle/kernel/physical_core.h

@@ -3,6 +3,7 @@
 
 #pragma once
 
+#include <condition_variable>
 #include <cstddef>
 #include <memory>
 #include <mutex>

src/core/hle/kernel/svc_version.h

@@ -35,11 +35,11 @@ constexpr inline u32 EncodeKernelVersion(u32 major, u32 minor) {
 }
 
 constexpr inline u32 GetKernelMajorVersion(u32 encoded) {
-    return std::bit_cast<decltype(KernelVersion::major_version)>(encoded).Value();
+    return decltype(KernelVersion::major_version)::ExtractValue(encoded);
 }
 
 constexpr inline u32 GetKernelMinorVersion(u32 encoded) {
-    return std::bit_cast<decltype(KernelVersion::minor_version)>(encoded).Value();
+    return decltype(KernelVersion::minor_version)::ExtractValue(encoded);
 }
 
 // Nintendo doesn't support programs targeting SVC versions < 3.0.

src/core/hle/service/hid/controllers/npad.cpp

@@ -758,12 +758,20 @@ Core::HID::NpadStyleTag Controller_NPad::GetSupportedStyleSet() const {
     return hid_core.GetSupportedStyleTag();
 }
 
-void Controller_NPad::SetSupportedNpadIdTypes(std::span<const u8> data) {
+Result Controller_NPad::SetSupportedNpadIdTypes(std::span<const u8> data) {
+    constexpr std::size_t max_number_npad_ids = 0xa;
     const auto length = data.size();
     ASSERT(length > 0 && (length % sizeof(u32)) == 0);
+    const std::size_t elements = length / sizeof(u32);
+
+    if (elements > max_number_npad_ids) {
+        return InvalidArraySize;
+    }
+
     supported_npad_id_types.clear();
-    supported_npad_id_types.resize(length / sizeof(u32));
+    supported_npad_id_types.resize(elements);
     std::memcpy(supported_npad_id_types.data(), data.data(), length);
+    return ResultSuccess;
 }
 
 void Controller_NPad::GetSupportedNpadIdTypes(u32* data, std::size_t max_length) {

src/core/hle/service/hid/controllers/npad.h

@@ -96,7 +96,7 @@ public:
     void SetSupportedStyleSet(Core::HID::NpadStyleTag style_set);
     Core::HID::NpadStyleTag GetSupportedStyleSet() const;
 
-    void SetSupportedNpadIdTypes(std::span<const u8> data);
+    Result SetSupportedNpadIdTypes(std::span<const u8> data);
     void GetSupportedNpadIdTypes(u32* data, std::size_t max_length);
     std::size_t GetSupportedNpadIdTypesSize() const;
 

src/core/hle/service/hid/errors.h

@@ -18,6 +18,7 @@ constexpr Result NpadIsDualJoycon{ErrorModule::HID, 601};
 constexpr Result NpadIsSameType{ErrorModule::HID, 602};
 constexpr Result InvalidNpadId{ErrorModule::HID, 709};
 constexpr Result NpadNotConnected{ErrorModule::HID, 710};
+constexpr Result InvalidArraySize{ErrorModule::HID, 715};
 constexpr Result InvalidPalmaHandle{ErrorModule::HID, 3302};
 
 } // namespace Service::HID

src/core/hle/service/hid/hid.cpp

@@ -1025,13 +1025,13 @@ void Hid::SetSupportedNpadIdType(Kernel::HLERequestContext& ctx) {
     IPC::RequestParser rp{ctx};
     const auto applet_resource_user_id{rp.Pop<u64>()};
 
-    applet_resource->GetController<Controller_NPad>(HidController::NPad)
-        .SetSupportedNpadIdTypes(ctx.ReadBuffer());
+    const auto result = applet_resource->GetController<Controller_NPad>(HidController::NPad)
+                            .SetSupportedNpadIdTypes(ctx.ReadBuffer());
 
     LOG_DEBUG(Service_HID, "called, applet_resource_user_id={}", applet_resource_user_id);
 
     IPC::ResponseBuilder rb{ctx, 2};
-    rb.Push(ResultSuccess);
+    rb.Push(result);
 }
 
 void Hid::ActivateNpad(Kernel::HLERequestContext& ctx) {

src/shader_recompiler/backend/glsl/glsl_emit_context.cpp

@@ -310,12 +310,6 @@ EmitContext::EmitContext(IR::Program& program, Bindings& bindings, const Profile
         if (runtime_info.force_early_z) {
             header += "layout(early_fragment_tests)in;";
         }
-        if (info.uses_sample_id) {
-            header += "in int gl_SampleID;";
-        }
-        if (info.stores_sample_mask) {
-            header += "out int gl_SampleMask[];";
-        }
         break;
     case Stage::Compute:
         stage_name = "cs";

src/tests/video_core/buffer_base.cpp

@@ -538,7 +538,7 @@ TEST_CASE("BufferBase: Cached write downloads") {
     int num = 0;
     buffer.ForEachDownloadRangeAndClear(c, WORD, [&](u64 offset, u64 size) { ++num; });
     buffer.ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
-    REQUIRE(num == 1);
+    REQUIRE(num == 0);
     REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
     REQUIRE(!buffer.IsRegionGpuModified(c + PAGE, PAGE));
     buffer.FlushCachedWrites();

src/video_core/buffer_cache/buffer_base.h

@@ -430,7 +430,7 @@ private:
         if (query_begin >= SizeBytes() || size < 0) {
             return;
         }
-        [[maybe_unused]] u64* const untracked_words = Array<Type::Untracked>();
+        u64* const untracked_words = Array<Type::Untracked>();
         u64* const state_words = Array<type>();
         const u64 query_end = query_begin + std::min(static_cast<u64>(size), SizeBytes());
         u64* const words_begin = state_words + query_begin / BYTES_PER_WORD;
@@ -483,7 +483,7 @@ private:
                 NotifyRasterizer<true>(word_index, current_bits, ~u64{0});
             }
             // Exclude CPU modified pages when visiting GPU pages
-            const u64 word = current_word;
+            const u64 word = current_word & ~(type == Type::GPU ? untracked_words[word_index] : 0);
             u64 page = page_begin;
             page_begin = 0;
 
@@ -531,7 +531,7 @@ private:
     [[nodiscard]] bool IsRegionModified(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
+        const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -539,7 +539,8 @@ private:
         const u64 page_limit = Common::DivCeil(offset + size, BYTES_PER_PAGE);
         u64 page_index = (offset / BYTES_PER_PAGE) % PAGES_PER_WORD;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index, page_index = 0) {
-            const u64 word = state_words[word_index];
+            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
+            const u64 word = state_words[word_index] & ~off_word;
             if (word == 0) {
                 continue;
             }
@@ -563,7 +564,7 @@ private:
     [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
+        const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -573,7 +574,8 @@ private:
         u64 begin = std::numeric_limits<u64>::max();
         u64 end = 0;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index) {
-            const u64 word = state_words[word_index];
+            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
+            const u64 word = state_words[word_index] & ~off_word;
             if (word == 0) {
                 continue;
             }

src/video_core/gpu.cpp

@@ -99,7 +99,7 @@ struct GPU::Impl {
 
     /// Signal the ending of command list.
     void OnCommandListEnd() {
-        rasterizer->ReleaseFences();
+        gpu_thread.OnCommandListEnd();
     }
 
     /// Request a host GPU memory flush from the CPU.

src/video_core/gpu_thread.cpp

@@ -40,6 +40,8 @@ static void RunThread(std::stop_token stop_token, Core::System& system,
             scheduler.Push(submit_list->channel, std::move(submit_list->entries));
         } else if (const auto* data = std::get_if<SwapBuffersCommand>(&next.data)) {
             renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr);
+        } else if (std::holds_alternative<OnCommandListEndCommand>(next.data)) {
+            rasterizer->ReleaseFences();
         } else if (std::holds_alternative<GPUTickCommand>(next.data)) {
             system.GPU().TickWork();
         } else if (const auto* flush = std::get_if<FlushRegionCommand>(&next.data)) {
@@ -108,6 +110,10 @@ void ThreadManager::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     rasterizer->OnCPUWrite(addr, size);
 }
 
+void ThreadManager::OnCommandListEnd() {
+    PushCommand(OnCommandListEndCommand());
+}
+
 u64 ThreadManager::PushCommand(CommandData&& command_data, bool block) {
     if (!is_async) {
         // In synchronous GPU mode, block the caller until the command has executed

src/video_core/gpu_thread.h

@@ -77,12 +77,16 @@ struct FlushAndInvalidateRegionCommand final {
     u64 size;
 };
 
+/// Command called within the gpu, to schedule actions after a command list end
+struct OnCommandListEndCommand final {};
+
 /// Command to make the gpu look into pending requests
 struct GPUTickCommand final {};
 
 using CommandData =
     std::variant<std::monostate, SubmitListCommand, SwapBuffersCommand, FlushRegionCommand,
-                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand, GPUTickCommand>;
+                 InvalidateRegionCommand, FlushAndInvalidateRegionCommand, OnCommandListEndCommand,
+                 GPUTickCommand>;
 
 struct CommandDataContainer {
     CommandDataContainer() = default;
@@ -130,6 +134,8 @@ public:
     /// Notify rasterizer that any caches of the specified region should be flushed and invalidated
     void FlushAndInvalidateRegion(VAddr addr, u64 size);
 
+    void OnCommandListEnd();
+
     void TickGPU();
 
 private:

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -548,31 +548,7 @@ void GraphicsPipeline::MakePipeline(VkRenderPass render_pass) {
     static_vector<VkVertexInputBindingDescription, 32> vertex_bindings;
     static_vector<VkVertexInputBindingDivisorDescriptionEXT, 32> vertex_binding_divisors;
     static_vector<VkVertexInputAttributeDescription, 32> vertex_attributes;
-    if (key.state.dynamic_vertex_input) {
-        const size_t num_vertex_arrays = std::min(
-            key.state.attributes.size(), static_cast<size_t>(device.GetMaxVertexInputBindings()));
-        for (size_t index = 0; index < num_vertex_arrays; ++index) {
-            const u32 type = key.state.DynamicAttributeType(index);
-            if (!stage_infos[0].loads.Generic(index) || type == 0) {
-                continue;
-            }
-            vertex_attributes.push_back({
-                .location = static_cast<u32>(index),
-                .binding = 0,
-                .format = type == 1   ? VK_FORMAT_R32_SFLOAT
-                          : type == 2 ? VK_FORMAT_R32_SINT
-                                      : VK_FORMAT_R32_UINT,
-                .offset = 0,
-            });
-        }
-        if (!vertex_attributes.empty()) {
-            vertex_bindings.push_back({
-                .binding = 0,
-                .stride = 4,
-                .inputRate = VK_VERTEX_INPUT_RATE_VERTEX,
-            });
-        }
-    } else {
+    if (!key.state.dynamic_vertex_input) {
         const size_t num_vertex_arrays = std::min(
             Maxwell::NumVertexArrays, static_cast<size_t>(device.GetMaxVertexInputBindings()));
         for (size_t index = 0; index < num_vertex_arrays; ++index) {