From 83a28dc251c287888d99e045a2682a89c8cd4efe Mon Sep 17 00:00:00 2001 From: lizzie Date: Fri, 16 Jan 2026 23:39:16 +0100 Subject: [common, core] remove uneeded memory indirection overhead at startup (#3306) for core stuff: just remove unique ptrs that dont need any pointer stability at all (afterall its an allocation within an allocation so yeah) for fibers: Main reasoning behind this is because virtualBuffer<> is stupidly fucking expensive and it also clutters my fstat view ALSO mmap is a syscall, syscalls are bad for performance or whatever ALSO std::vector<> is better suited for handling this kind of "fixed size thing where its like big but not THAT big" (512 KiB isn't going to kill your memory usage for each fiber...) for core.cpp stuff - inlines stuff into std::optional<> as opposed to std::unique_ptr<> (because yknow, we are making the Impl from an unique_ptr, allocating within an allocation is unnecessary) - reorganizes the structures a bit so padding doesnt screw us up (it's not perfect but eh saves a measly 44 bytes) - removes unused/dead code - uses std::vector<> instead of std::deque<> no perf impact expected, maybe some initialisation boost but very minimal impact nonethless lto gets rid of most calls anyways - the heavy issue is with shared_ptr and the cache coherency from the atomics... but i clumped them together because well, they kinda do not suffer from cache coherency - hopefully not a mistake this balloons the size of Impl to about 1.67 MB - which is fine because we throw it in the stack anyways REST OF INTERFACES: most of them ballooned in size as well, but overhead is ok since its an allocation within an alloc, no stack is used (when it comes to storing these i mean) Signed-off-by: lizzie lizzie@eden-emu.dev Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3306 Reviewed-by: CamilleLaVey Reviewed-by: MaranBr Co-authored-by: lizzie Co-committed-by: lizzie --- src/core/hle/kernel/kernel.cpp | 119 ++++++++++++++++++++--------------------- 1 file changed, 57 insertions(+), 62 deletions(-) (limited to 'src/core/hle/kernel/kernel.cpp') diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp index 062387a29b..6986a98e35 100644 --- a/src/core/hle/kernel/kernel.cpp +++ b/src/core/hle/kernel/kernel.cpp @@ -88,11 +88,11 @@ struct KernelCore::Impl { } void Initialize(KernelCore& kernel) { - hardware_timer = std::make_unique(kernel); + hardware_timer.emplace(kernel); hardware_timer->Initialize(); - global_object_list_container = std::make_unique(kernel); - global_scheduler_context = std::make_unique(kernel); + global_object_list_container.emplace(kernel); + global_scheduler_context.emplace(kernel); // Derive the initial memory layout from the emulated board Init::InitializeSlabResourceCounts(kernel); @@ -212,10 +212,9 @@ struct KernelCore::Impl { void InitializePhysicalCores() { for (u32 i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) { - const s32 core{static_cast(i)}; - - schedulers[i] = std::make_unique(system.Kernel()); - cores[i] = std::make_unique(system.Kernel(), i); + auto const core = s32(i); + schedulers[i].emplace(system.Kernel()); + cores[i].emplace(system.Kernel(), i); auto* main_thread{Kernel::KThread::Create(system.Kernel())}; main_thread->SetCurrentCore(core); @@ -280,57 +279,56 @@ struct KernelCore::Impl { size -= rc_size; // Initialize the resource managers' shared page manager. - resource_manager_page_manager = std::make_unique(); + resource_manager_page_manager.emplace(); resource_manager_page_manager->Initialize(address, size, std::max(PageSize, KPageBufferSlabHeap::BufferSize)); // Initialize the KPageBuffer slab heap. page_buffer_slab_heap.Initialize(system); // Initialize the fixed-size slabheaps. - app_memory_block_heap = std::make_unique(); - sys_memory_block_heap = std::make_unique(); - block_info_heap = std::make_unique(); - app_memory_block_heap->Initialize(resource_manager_page_manager.get(), ApplicationMemoryBlockSlabHeapSize); - sys_memory_block_heap->Initialize(resource_manager_page_manager.get(), SystemMemoryBlockSlabHeapSize); - block_info_heap->Initialize(resource_manager_page_manager.get(), BlockInfoSlabHeapSize); + app_memory_block_heap.emplace(); + sys_memory_block_heap.emplace(); + block_info_heap.emplace(); + app_memory_block_heap->Initialize(std::addressof(*resource_manager_page_manager), ApplicationMemoryBlockSlabHeapSize); + sys_memory_block_heap->Initialize(std::addressof(*resource_manager_page_manager), SystemMemoryBlockSlabHeapSize); + block_info_heap->Initialize(std::addressof(*resource_manager_page_manager), BlockInfoSlabHeapSize); // Reserve all but a fixed number of remaining pages for the page table heap. const size_t num_pt_pages = resource_manager_page_manager->GetCount() - resource_manager_page_manager->GetUsed() - ReservedDynamicPageCount; - page_table_heap = std::make_unique(); + page_table_heap.emplace(); // TODO(bunnei): Pass in address once we support kernel virtual memory allocations. page_table_heap->Initialize( - resource_manager_page_manager.get(), num_pt_pages, + std::addressof(*resource_manager_page_manager), num_pt_pages, /*GetPointer(address + size)*/ nullptr); // Setup the slab managers. KDynamicPageManager* const app_dynamic_page_manager = nullptr; KDynamicPageManager* const sys_dynamic_page_manager = /*KTargetSystem::IsDynamicResourceLimitsEnabled()*/ true - ? resource_manager_page_manager.get() - : nullptr; - app_memory_block_manager = std::make_unique(); - sys_memory_block_manager = std::make_unique(); - app_block_info_manager = std::make_unique(); - sys_block_info_manager = std::make_unique(); - app_page_table_manager = std::make_unique(); - sys_page_table_manager = std::make_unique(); + ? std::addressof(*resource_manager_page_manager) : nullptr; + app_memory_block_manager.emplace(); + sys_memory_block_manager.emplace(); + app_block_info_manager.emplace(); + sys_block_info_manager.emplace(); + app_page_table_manager.emplace(); + sys_page_table_manager.emplace(); - app_memory_block_manager->Initialize(app_dynamic_page_manager, app_memory_block_heap.get()); - sys_memory_block_manager->Initialize(sys_dynamic_page_manager, sys_memory_block_heap.get()); + app_memory_block_manager->Initialize(app_dynamic_page_manager, std::addressof(*app_memory_block_heap)); + sys_memory_block_manager->Initialize(sys_dynamic_page_manager, std::addressof(*sys_memory_block_heap)); - app_block_info_manager->Initialize(app_dynamic_page_manager, block_info_heap.get()); - sys_block_info_manager->Initialize(sys_dynamic_page_manager, block_info_heap.get()); + app_block_info_manager->Initialize(app_dynamic_page_manager, std::addressof(*block_info_heap)); + sys_block_info_manager->Initialize(sys_dynamic_page_manager, std::addressof(*block_info_heap)); - app_page_table_manager->Initialize(app_dynamic_page_manager, page_table_heap.get()); - sys_page_table_manager->Initialize(sys_dynamic_page_manager, page_table_heap.get()); + app_page_table_manager->Initialize(app_dynamic_page_manager, std::addressof(*page_table_heap)); + sys_page_table_manager->Initialize(sys_dynamic_page_manager, std::addressof(*page_table_heap)); // Check that we have the correct number of dynamic pages available. ASSERT(resource_manager_page_manager->GetCount() - resource_manager_page_manager->GetUsed() == ReservedDynamicPageCount); // Create the system page table managers. - app_system_resource = std::make_unique(kernel); - sys_system_resource = std::make_unique(kernel); + app_system_resource.emplace(kernel); + sys_system_resource.emplace(kernel); KAutoObject::Create(std::addressof(*app_system_resource)); KAutoObject::Create(std::addressof(*sys_system_resource)); @@ -349,7 +347,7 @@ struct KernelCore::Impl { } void InitializeGlobalData(KernelCore& kernel) { - object_name_global_data = std::make_unique(kernel); + object_name_global_data.emplace(kernel); } void MakeApplicationProcess(KProcess* process) { @@ -431,7 +429,7 @@ struct KernelCore::Impl { } void DeriveInitialMemoryLayout() { - memory_layout = std::make_unique(); + memory_layout.emplace(); // Insert the root region for the virtual memory tree, from which all other regions will // derive. @@ -726,7 +724,7 @@ struct KernelCore::Impl { void InitializeMemoryLayout() { // Initialize the memory manager. - memory_manager = std::make_unique(system); + memory_manager.emplace(system); const auto& management_region = memory_layout->GetPoolManagementRegion(); ASSERT(management_region.GetEndAddress() != 0); memory_manager->Initialize(management_region.GetAddress(), management_region.GetSize()); @@ -774,8 +772,8 @@ struct KernelCore::Impl { std::mutex process_list_lock; std::vector process_list; KProcess* application_process{}; - std::unique_ptr global_scheduler_context; - std::unique_ptr hardware_timer; + std::optional global_scheduler_context; + std::optional hardware_timer; Init::KSlabResourceCounts slab_resource_counts{}; KResourceLimit* system_resource_limit{}; @@ -784,9 +782,9 @@ struct KernelCore::Impl { std::shared_ptr preemption_event; - std::unique_ptr global_object_list_container; + std::optional global_object_list_container; - std::unique_ptr object_name_global_data; + std::optional object_name_global_data; std::unordered_set registered_objects; std::unordered_set registered_in_use_objects; @@ -794,28 +792,28 @@ struct KernelCore::Impl { std::mutex server_lock; std::vector> server_managers; - std::array, Core::Hardware::NUM_CPU_CORES> cores; + std::array, Core::Hardware::NUM_CPU_CORES> cores; // Next host thead ID to use, 0-3 IDs represent core threads, >3 represent others std::atomic next_host_thread_id{Core::Hardware::NUM_CPU_CORES}; // Kernel memory management - std::unique_ptr memory_manager; + std::optional memory_manager; // Resource managers - std::unique_ptr resource_manager_page_manager; - std::unique_ptr page_table_heap; - std::unique_ptr app_memory_block_heap; - std::unique_ptr sys_memory_block_heap; - std::unique_ptr block_info_heap; - std::unique_ptr app_page_table_manager; - std::unique_ptr sys_page_table_manager; - std::unique_ptr app_memory_block_manager; - std::unique_ptr sys_memory_block_manager; - std::unique_ptr app_block_info_manager; - std::unique_ptr sys_block_info_manager; - std::unique_ptr app_system_resource; - std::unique_ptr sys_system_resource; + std::optional resource_manager_page_manager; + std::optional page_table_heap; + std::optional app_memory_block_heap; + std::optional sys_memory_block_heap; + std::optional block_info_heap; + std::optional app_page_table_manager; + std::optional sys_page_table_manager; + std::optional app_memory_block_manager; + std::optional sys_memory_block_manager; + std::optional app_block_info_manager; + std::optional sys_block_info_manager; + std::optional app_system_resource; + std::optional sys_system_resource; // Shared memory for services Kernel::KSharedMemory* hid_shared_mem{}; @@ -825,10 +823,10 @@ struct KernelCore::Impl { Kernel::KSharedMemory* hidbus_shared_mem{}; // Memory layout - std::unique_ptr memory_layout; + std::optional memory_layout; std::array shutdown_threads{}; - std::array, Core::Hardware::NUM_CPU_CORES> schedulers{}; + std::array, Core::Hardware::NUM_CPU_CORES> schedulers{}; bool is_multicore{}; std::atomic_bool is_shutting_down{}; @@ -948,12 +946,9 @@ const Kernel::PhysicalCore& KernelCore::CurrentPhysicalCore() const { } Kernel::KScheduler* KernelCore::CurrentScheduler() { - const u32 core_id = impl->GetCurrentHostThreadID(); - if (core_id >= Core::Hardware::NUM_CPU_CORES) { - // This is expected when called from not a guest thread - return {}; - } - return impl->schedulers[core_id].get(); + if (auto const core_id = impl->GetCurrentHostThreadID(); core_id < Core::Hardware::NUM_CPU_CORES) + return std::addressof(*impl->schedulers[core_id]); + return {}; // This is expected when called from not a guest thread } Kernel::KHardwareTimer& KernelCore::HardwareTimer() { -- cgit v1.2.3