diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt index ec2759edd5..0a40daa486 100644 --- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt +++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/BooleanSetting.kt @@ -13,6 +13,7 @@ enum class BooleanSetting(override val key: String) : AbstractBooleanSetting { CORE_SYNC_CORE_SPEED("sync_core_speed"), RENDERER_USE_SPEED_LIMIT("use_speed_limit"), USE_DOCKED_MODE("use_docked_mode"), + USE_AUTO_STUB("use_auto_stub"), RENDERER_USE_DISK_SHADER_CACHE("use_disk_shader_cache"), RENDERER_FORCE_MAX_CLOCK("force_max_clock"), RENDERER_ASYNCHRONOUS_SHADERS("use_asynchronous_shaders"), diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt index 7ccc6500c2..52f63ad3d1 100644 --- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt +++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/model/view/SettingsItem.kt @@ -273,7 +273,13 @@ abstract class SettingsItem( descriptionId = R.string.use_docked_mode_description ) ) - + put( + SwitchSetting( + BooleanSetting.USE_AUTO_STUB, + titleId = R.string.use_auto_stub, + descriptionId = R.string.use_auto_stub_description + ) + ) put( SwitchSetting( syncCoreSpeedSetting, @@ -354,6 +360,7 @@ abstract class SettingsItem( valuesId = R.array.vramUsageMethodValues ) ) + put( SingleChoiceSetting( IntSetting.RENDERER_RESOLUTION, diff --git a/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt b/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt index 4a33b176d7..8f136790a6 100644 --- a/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt +++ b/src/android/app/src/main/java/org/yuzu/yuzu_emu/features/settings/ui/SettingsFragmentPresenter.kt @@ -261,6 +261,7 @@ class SettingsFragmentPresenter( add(BooleanSetting.RENDERER_USE_SPEED_LIMIT.key) add(ShortSetting.RENDERER_SPEED_LIMIT.key) add(BooleanSetting.USE_DOCKED_MODE.key) + add(BooleanSetting.USE_AUTO_STUB.key) add(IntSetting.REGION_INDEX.key) add(IntSetting.LANGUAGE_INDEX.key) add(BooleanSetting.USE_CUSTOM_RTC.key) diff --git a/src/android/app/src/main/res/values/strings.xml b/src/android/app/src/main/res/values/strings.xml index 28f18e1596..2987ea91fc 100644 --- a/src/android/app/src/main/res/values/strings.xml +++ b/src/android/app/src/main/res/values/strings.xml @@ -256,6 +256,8 @@ Device name Docked Mode Increases resolution, decreasing performance. Handheld Mode is used when disabled, lowering resolution and increasing performance. + Use Auto Stub + Automatically stub missing services and functions. This may improve compatibility but can cause crashes and stability issues. Emulated region Emulated language Select RTC date diff --git a/src/common/settings.h b/src/common/settings.h index ec627b744d..7316b27e2d 100644 --- a/src/common/settings.h +++ b/src/common/settings.h @@ -622,7 +622,7 @@ struct Values { linkage, false, "extended_logging", Category::Debugging, Specialization::Default, false}; Setting use_debug_asserts{linkage, false, "use_debug_asserts", Category::Debugging}; Setting use_auto_stub{ - linkage, false, "use_auto_stub", Category::Debugging, Specialization::Default, false}; + linkage, false, "use_auto_stub", Category::Debugging}; Setting enable_all_controllers{linkage, false, "enable_all_controllers", Category::Debugging}; Setting perform_vulkan_check{linkage, true, "perform_vulkan_check", Category::Debugging}; @@ -636,7 +636,7 @@ struct Values { Category::Network}; // WebService - Setting web_api_url{linkage, "https://api.ynet-fun.xyz", "web_api_url", + Setting web_api_url{linkage, "api.ynet-fun.xyz", "web_api_url", Category::WebService}; Setting yuzu_username{linkage, std::string(), "yuzu_username", Category::WebService}; diff --git a/src/core/hle/kernel/k_process.h b/src/core/hle/kernel/k_process.h index ab1358a129..df3e540dc2 100644 --- a/src/core/hle/kernel/k_process.h +++ b/src/core/hle/kernel/k_process.h @@ -84,6 +84,7 @@ private: std::array m_entropy{}; bool m_is_signaled{}; bool m_is_initialized{}; + u32 m_pointer_buffer_size = 0x8000; // Default pointer buffer size (can be game-specific later) bool m_is_application{}; bool m_is_default_application_system_resource{}; bool m_is_hbl{}; @@ -239,6 +240,14 @@ public: m_is_suspended = suspended; } + u32 GetPointerBufferSize() const { + return m_pointer_buffer_size; + } + + void SetPointerBufferSize(u32 size) { + m_pointer_buffer_size = size; + } + Result Terminate(); bool IsTerminated() const { diff --git a/src/core/hle/service/am/service/common_state_getter.cpp b/src/core/hle/service/am/service/common_state_getter.cpp index c445e5e776..c179102f76 100644 --- a/src/core/hle/service/am/service/common_state_getter.cpp +++ b/src/core/hle/service/am/service/common_state_getter.cpp @@ -38,7 +38,7 @@ ICommonStateGetter::ICommonStateGetter(Core::System& system_, std::shared_ptr, "GetReaderLockAccessorEx"}, {32, D<&ICommonStateGetter::GetWriterLockAccessorEx>, "GetWriterLockAccessorEx"}, - {40, D<&ICommonStateGetter::GetCradleFwVersion>, "GetCradleFwVersion"}, + {40, nullptr, "GetCradleFwVersion"}, {50, D<&ICommonStateGetter::IsVrModeEnabled>, "IsVrModeEnabled"}, {51, D<&ICommonStateGetter::SetVrModeEnabled>, "SetVrModeEnabled"}, {52, D<&ICommonStateGetter::SetLcdBacklighOffEnabled>, "SetLcdBacklighOffEnabled"}, @@ -172,17 +172,6 @@ Result ICommonStateGetter::GetBootMode(Out out_boot_mode) { R_SUCCEED(); } -Result ICommonStateGetter::GetCradleFwVersion(OutArray out_version) { - LOG_DEBUG(Service_AM, "(STUBBED) called"); - - out_version[0] = 0; - out_version[1] = 0; - out_version[2] = 0; - out_version[3] = 0; - - R_SUCCEED(); -} - Result ICommonStateGetter::IsVrModeEnabled(Out out_is_vr_mode_enabled) { LOG_DEBUG(Service_AM, "called"); diff --git a/src/core/hle/service/am/service/common_state_getter.h b/src/core/hle/service/am/service/common_state_getter.h index a326e11053..0cafbfe176 100644 --- a/src/core/hle/service/am/service/common_state_getter.h +++ b/src/core/hle/service/am/service/common_state_getter.h @@ -39,7 +39,6 @@ private: Result GetHdcpAuthenticationStateChangeEvent(OutCopyHandle out_event); Result GetOperationMode(Out out_operation_mode); Result GetPerformanceMode(Out out_performance_mode); - Result GetCradleFwVersion(OutArray out_version); Result GetBootMode(Out out_boot_mode); Result IsVrModeEnabled(Out out_is_vr_mode_enabled); Result SetVrModeEnabled(bool is_vr_mode_enabled); diff --git a/src/core/hle/service/sm/sm_controller.cpp b/src/core/hle/service/sm/sm_controller.cpp index 7f0fb91d04..9e25eae4d4 100644 --- a/src/core/hle/service/sm/sm_controller.cpp +++ b/src/core/hle/service/sm/sm_controller.cpp @@ -68,13 +68,46 @@ void Controller::CloneCurrentObjectEx(HLERequestContext& ctx) { } void Controller::QueryPointerBufferSize(HLERequestContext& ctx) { - LOG_WARNING(Service, "(STUBBED) called"); + LOG_DEBUG(Service, "called"); + + auto* process = Kernel::GetCurrentProcessPointer(kernel); + ASSERT(process != nullptr); + + u32 buffer_size = process->GetPointerBufferSize(); + if (buffer_size > std::numeric_limits::max()) { + LOG_WARNING(Service, "Pointer buffer size exceeds u16 max, clamping"); + buffer_size = std::numeric_limits::max(); + } IPC::ResponseBuilder rb{ctx, 3}; rb.Push(ResultSuccess); - rb.Push(0x8000); + rb.Push(static_cast(buffer_size)); } +void Controller::SetPointerBufferSize(HLERequestContext& ctx) { + LOG_DEBUG(Service, "called"); + + auto* process = Kernel::GetCurrentProcessPointer(kernel); + ASSERT(process != nullptr); + + IPC::RequestParser rp{ctx}; + + u32 requested_size = rp.PopRaw(); + + if (requested_size > std::numeric_limits::max()) { + LOG_WARNING(Service, "Requested pointer buffer size too large, clamping to 0xFFFF"); + requested_size = std::numeric_limits::max(); + } + + process->SetPointerBufferSize(requested_size); + + LOG_INFO(Service, "Pointer buffer size dynamically updated to {:#x} bytes by process", requested_size); + + IPC::ResponseBuilder rb{ctx, 2}; + rb.Push(ResultSuccess); +} + + // https://switchbrew.org/wiki/IPC_Marshalling Controller::Controller(Core::System& system_) : ServiceFramework{system_, "IpcController"} { static const FunctionInfo functions[] = { @@ -83,6 +116,7 @@ Controller::Controller(Core::System& system_) : ServiceFramework{system_, "IpcCo {2, &Controller::CloneCurrentObject, "CloneCurrentObject"}, {3, &Controller::QueryPointerBufferSize, "QueryPointerBufferSize"}, {4, &Controller::CloneCurrentObjectEx, "CloneCurrentObjectEx"}, + {5, &Controller::SetPointerBufferSize, "SetPointerBufferSize"}, }; RegisterHandlers(functions); } diff --git a/src/core/hle/service/sm/sm_controller.h b/src/core/hle/service/sm/sm_controller.h index 4e748b36d9..f7e529a269 100644 --- a/src/core/hle/service/sm/sm_controller.h +++ b/src/core/hle/service/sm/sm_controller.h @@ -21,6 +21,7 @@ private: void CloneCurrentObject(HLERequestContext& ctx); void CloneCurrentObjectEx(HLERequestContext& ctx); void QueryPointerBufferSize(HLERequestContext& ctx); + void SetPointerBufferSize(HLERequestContext& ctx); }; } // namespace Service::SM diff --git a/src/core/loader/nca.cpp b/src/core/loader/nca.cpp index de27ec49e2..4a87ab53e7 100644 --- a/src/core/loader/nca.cpp +++ b/src/core/loader/nca.cpp @@ -15,9 +15,20 @@ #include "core/loader/deconstructed_rom_directory.h" #include "core/loader/nca.h" #include "mbedtls/sha256.h" +#include "common/literals.h" namespace Loader { +static u32 CalculatePointerBufferSize(size_t heap_size) { + if (heap_size > 1073741824) { // Games with 1 GiB + return 0x10000; + } else if (heap_size > 536870912) { // Games with 512 MiB + return 0xC000; + } else { + return 0x8000; // Default for all other games + } +} + AppLoader_NCA::AppLoader_NCA(FileSys::VirtualFile file_) : AppLoader(std::move(file_)), nca(std::make_unique(file)) {} @@ -52,8 +63,6 @@ AppLoader_NCA::LoadResult AppLoader_NCA::Load(Kernel::KProcess& process, Core::S if (exefs == nullptr) { LOG_INFO(Loader, "No ExeFS found in NCA, looking for ExeFS from update"); - // This NCA may be a sparse base of an installed title. - // Try to fetch the ExeFS from the installed update. const auto& installed = system.GetContentProvider(); const auto update_nca = installed.GetEntry(FileSys::GetUpdateTitleID(nca->GetTitleId()), FileSys::ContentRecordType::Program); @@ -69,11 +78,37 @@ AppLoader_NCA::LoadResult AppLoader_NCA::Load(Kernel::KProcess& process, Core::S directory_loader = std::make_unique(exefs, true); + // Read heap size from main.npdm in ExeFS + u64 heap_size = 0; + + if (exefs) { + const auto npdm_file = exefs->GetFile("main.npdm"); + if (npdm_file) { + auto npdm_data = npdm_file->ReadAllBytes(); + if (npdm_data.size() >= 0x30) { + heap_size = *reinterpret_cast(&npdm_data[0x28]); + LOG_INFO(Loader, "Read heap size {:#x} bytes from main.npdm", heap_size); + } else { + LOG_WARNING(Loader, "main.npdm too small to read heap size!"); + } + } else { + LOG_WARNING(Loader, "No main.npdm found in ExeFS!"); + } + } + + // Set pointer buffer size based on heap size + process.SetPointerBufferSize(CalculatePointerBufferSize(heap_size)); + + // Load modules const auto load_result = directory_loader->Load(process, system); if (load_result.first != ResultStatus::Success) { return load_result; } + LOG_INFO(Loader, "Set pointer buffer size to {:#x} bytes for ProgramID {:#018x} (Heap size: {:#x})", + process.GetPointerBufferSize(), nca->GetTitleId(), heap_size); + + // Register the process in the file system controller system.GetFileSystemController().RegisterProcess( process.GetProcessId(), nca->GetTitleId(), std::make_shared(*this, system.GetContentProvider(), diff --git a/src/core/memory.cpp b/src/core/memory.cpp index 6f7703fced..b033858bf8 100644 --- a/src/core/memory.cpp +++ b/src/core/memory.cpp @@ -6,6 +6,8 @@ #include #include #include +#include +#include #include "common/assert.h" #include "common/atomic_ops.h" @@ -32,6 +34,105 @@ namespace Core::Memory { namespace { +inline void FastMemcpy(void* dst, const void* src, std::size_t size) { + // Fast path for small copies + switch (size) { + case 1: + *static_cast(dst) = *static_cast(src); + break; + case 2: + *static_cast(dst) = *static_cast(src); + break; + case 4: + *static_cast(dst) = *static_cast(src); + break; + case 8: + *static_cast(dst) = *static_cast(src); + break; + case 16: { + // Optimize for 16-byte copy (common case for SIMD registers) + const u64* src_64 = static_cast(src); + u64* dst_64 = static_cast(dst); + dst_64[0] = src_64[0]; + dst_64[1] = src_64[1]; + break; + } + case 32: { + // Optimize for 32-byte copy + const u64* src_64 = static_cast(src); + u64* dst_64 = static_cast(dst); + dst_64[0] = src_64[0]; + dst_64[1] = src_64[1]; + dst_64[2] = src_64[2]; + dst_64[3] = src_64[3]; + break; + } + case 64: { + // Optimize for 64-byte copy + const u64* src_64 = static_cast(src); + u64* dst_64 = static_cast(dst); + dst_64[0] = src_64[0]; + dst_64[1] = src_64[1]; + dst_64[2] = src_64[2]; + dst_64[3] = src_64[3]; + dst_64[4] = src_64[4]; + dst_64[5] = src_64[5]; + dst_64[6] = src_64[6]; + dst_64[7] = src_64[7]; + break; + } + default: + // For larger sizes, use standard memcpy which is usually optimized by the compiler + std::memcpy(dst, src, size); + break; + } +} + +inline void FastMemset(void* dst, int value, std::size_t size) { + // Fast path for small fills + switch (size) { + case 1: + *static_cast(dst) = static_cast(value); + break; + case 2: + *static_cast(dst) = static_cast(value); + break; + case 4: + *static_cast(dst) = static_cast(value); + break; + case 8: + *static_cast(dst) = static_cast(value); + break; + case 16: { + // Optimize for 16-byte fill (common case for SIMD registers) + u64* dst_64 = static_cast(dst); + const u64 val64 = static_cast(value) * 0x0101010101010101ULL; + dst_64[0] = val64; + dst_64[1] = val64; + break; + } + default: + if (size <= 128 && value == 0) { + // Fast path for small zero-fills + u8* dst_bytes = static_cast(dst); + for (std::size_t i = 0; i < size; i += 8) { + if (i + 8 <= size) { + *reinterpret_cast(dst_bytes + i) = 0; + } else { + // Handle remaining bytes (less than 8) + for (std::size_t j = i; j < size; j++) { + dst_bytes[j] = 0; + } + } + } + } else { + // For larger sizes, use standard memset which is usually optimized by the compiler + std::memset(dst, value, size); + } + break; + } +} + bool AddressSpaceContains(const Common::PageTable& table, const Common::ProcessAddress addr, const std::size_t size) { const Common::ProcessAddress max_addr = 1ULL << table.GetAddressSpaceBits(); @@ -44,7 +145,11 @@ bool AddressSpaceContains(const Common::PageTable& table, const Common::ProcessA // from outside classes. This also allows modification to the internals of the memory // subsystem without needing to rebuild all files that make use of the memory interface. struct Memory::Impl { - explicit Impl(Core::System& system_) : system{system_} {} + explicit Impl(Core::System& system_) : system{system_} { + // Initialize thread count based on available cores for parallel memory operations + const unsigned int hw_concurrency = std::thread::hardware_concurrency(); + thread_count = std::max(2u, std::min(hw_concurrency, 8u)); // Limit to 8 threads max + } void SetCurrentPageTable(Kernel::KProcess& process) { current_page_table = &process.GetPageTable().GetImpl(); @@ -308,26 +413,70 @@ struct Memory::Impl { LOG_ERROR(HW_Memory, "Unmapped ReadBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})", GetInteger(current_vaddr), GetInteger(src_addr), size); - std::memset(dest_buffer, 0, copy_amount); + FastMemset(dest_buffer, 0, copy_amount); }, [&](const std::size_t copy_amount, const u8* const src_ptr) { - std::memcpy(dest_buffer, src_ptr, copy_amount); + FastMemcpy(dest_buffer, src_ptr, copy_amount); }, [&](const Common::ProcessAddress current_vaddr, const std::size_t copy_amount, const u8* const host_ptr) { if constexpr (!UNSAFE) { HandleRasterizerDownload(GetInteger(current_vaddr), copy_amount); } - std::memcpy(dest_buffer, host_ptr, copy_amount); + FastMemcpy(dest_buffer, host_ptr, copy_amount); }, [&](const std::size_t copy_amount) { dest_buffer = static_cast(dest_buffer) + copy_amount; }); } + bool ReadBlockParallel(const Common::ProcessAddress src_addr, void* dest_buffer, + const std::size_t size) { + // Calculate chunk size based on thread count + const size_t chunk_size = (size + thread_count - 1) / thread_count; + + // Create threads for parallel processing + std::vector threads; + threads.reserve(thread_count); + + // Create a vector to store the results of each thread + std::vector results(thread_count, true); + + // Split the work among threads + for (unsigned int i = 0; i < thread_count; ++i) { + const size_t offset = i * chunk_size; + if (offset >= size) { + break; + } + + const size_t current_chunk_size = std::min(chunk_size, size - offset); + const Common::ProcessAddress current_addr = src_addr + offset; + void* current_dest = static_cast(dest_buffer) + offset; + + // Launch thread + threads.emplace_back([this, i, current_addr, current_dest, current_chunk_size, &results] { + results[i] = ReadBlockImpl(current_addr, current_dest, current_chunk_size); + }); + } + + // Wait for all threads to complete + for (auto& thread : threads) { + thread.join(); + } + + // Check if all operations succeeded + return std::all_of(results.begin(), results.end(), [](bool result) { return result; }); + } + bool ReadBlock(const Common::ProcessAddress src_addr, void* dest_buffer, const std::size_t size) { - return ReadBlockImpl(src_addr, dest_buffer, size); + // For small reads, use the regular implementation + if (size < PARALLEL_THRESHOLD) { + return ReadBlockImpl(src_addr, dest_buffer, size); + } + + // For large reads, use parallel implementation + return ReadBlockParallel(src_addr, dest_buffer, size); } bool ReadBlockUnsafe(const Common::ProcessAddress src_addr, void* dest_buffer, @@ -363,23 +512,67 @@ struct Memory::Impl { GetInteger(current_vaddr), GetInteger(dest_addr), size); }, [&](const std::size_t copy_amount, u8* const dest_ptr) { - std::memcpy(dest_ptr, src_buffer, copy_amount); + FastMemcpy(dest_ptr, src_buffer, copy_amount); }, [&](const Common::ProcessAddress current_vaddr, const std::size_t copy_amount, u8* const host_ptr) { if constexpr (!UNSAFE) { HandleRasterizerWrite(GetInteger(current_vaddr), copy_amount); } - std::memcpy(host_ptr, src_buffer, copy_amount); + FastMemcpy(host_ptr, src_buffer, copy_amount); }, [&](const std::size_t copy_amount) { src_buffer = static_cast(src_buffer) + copy_amount; }); } + bool WriteBlockParallel(const Common::ProcessAddress dest_addr, const void* src_buffer, + const std::size_t size) { + // Calculate chunk size based on thread count + const size_t chunk_size = (size + thread_count - 1) / thread_count; + + // Create threads for parallel processing + std::vector threads; + threads.reserve(thread_count); + + // Create a vector to store the results of each thread + std::vector results(thread_count, true); + + // Split the work among threads + for (unsigned int i = 0; i < thread_count; ++i) { + const size_t offset = i * chunk_size; + if (offset >= size) { + break; + } + + const size_t current_chunk_size = std::min(chunk_size, size - offset); + const Common::ProcessAddress current_addr = dest_addr + offset; + const void* current_src = static_cast(src_buffer) + offset; + + // Launch thread + threads.emplace_back([this, i, current_addr, current_src, current_chunk_size, &results] { + results[i] = WriteBlockImpl(current_addr, current_src, current_chunk_size); + }); + } + + // Wait for all threads to complete + for (auto& thread : threads) { + thread.join(); + } + + // Check if all operations succeeded + return std::all_of(results.begin(), results.end(), [](bool result) { return result; }); + } + bool WriteBlock(const Common::ProcessAddress dest_addr, const void* src_buffer, const std::size_t size) { - return WriteBlockImpl(dest_addr, src_buffer, size); + // For small writes, use the regular implementation + if (size < PARALLEL_THRESHOLD) { + return WriteBlockImpl(dest_addr, src_buffer, size); + } + + // For large writes, use parallel implementation + return WriteBlockParallel(dest_addr, src_buffer, size); } bool WriteBlockUnsafe(const Common::ProcessAddress dest_addr, const void* src_buffer, @@ -397,12 +590,12 @@ struct Memory::Impl { GetInteger(current_vaddr), GetInteger(dest_addr), size); }, [](const std::size_t copy_amount, u8* const dest_ptr) { - std::memset(dest_ptr, 0, copy_amount); + FastMemset(dest_ptr, 0, copy_amount); }, [&](const Common::ProcessAddress current_vaddr, const std::size_t copy_amount, u8* const host_ptr) { HandleRasterizerWrite(GetInteger(current_vaddr), copy_amount); - std::memset(host_ptr, 0, copy_amount); + FastMemset(host_ptr, 0, copy_amount); }, [](const std::size_t copy_amount) {}); } @@ -733,16 +926,71 @@ struct Memory::Impl { */ template T Read(Common::ProcessAddress vaddr) { + // Fast path for aligned reads of common sizes + const u64 addr = GetInteger(vaddr); + if constexpr (std::is_same_v || std::is_same_v) { + // 8-bit reads are always aligned + const u8* const ptr = GetPointerImpl( + addr, + [addr]() { + LOG_ERROR(HW_Memory, "Unmapped Read8 @ 0x{:016X}", addr); + }, + [&]() { HandleRasterizerDownload(addr, sizeof(T)); }); + if (ptr) { + return static_cast(*ptr); + } + return 0; + } else if constexpr (std::is_same_v || std::is_same_v) { + // Check alignment for 16-bit reads + if ((addr & 1) == 0) { + const u8* const ptr = GetPointerImpl( + addr, + [addr]() { + LOG_ERROR(HW_Memory, "Unmapped Read16 @ 0x{:016X}", addr); + }, + [&]() { HandleRasterizerDownload(addr, sizeof(T)); }); + if (ptr) { + return static_cast(*reinterpret_cast(ptr)); + } + } + } else if constexpr (std::is_same_v || std::is_same_v) { + // Check alignment for 32-bit reads + if ((addr & 3) == 0) { + const u8* const ptr = GetPointerImpl( + addr, + [addr]() { + LOG_ERROR(HW_Memory, "Unmapped Read32 @ 0x{:016X}", addr); + }, + [&]() { HandleRasterizerDownload(addr, sizeof(T)); }); + if (ptr) { + return static_cast(*reinterpret_cast(ptr)); + } + } + } else if constexpr (std::is_same_v || std::is_same_v) { + // Check alignment for 64-bit reads + if ((addr & 7) == 0) { + const u8* const ptr = GetPointerImpl( + addr, + [addr]() { + LOG_ERROR(HW_Memory, "Unmapped Read64 @ 0x{:016X}", addr); + }, + [&]() { HandleRasterizerDownload(addr, sizeof(T)); }); + if (ptr) { + return static_cast(*reinterpret_cast(ptr)); + } + } + } + + // Fall back to the general case for other types or unaligned access T result = 0; const u8* const ptr = GetPointerImpl( - GetInteger(vaddr), - [vaddr]() { - LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8, - GetInteger(vaddr)); + addr, + [addr]() { + LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8, addr); }, - [&]() { HandleRasterizerDownload(GetInteger(vaddr), sizeof(T)); }); + [&]() { HandleRasterizerDownload(addr, sizeof(T)); }); if (ptr) { - std::memcpy(&result, ptr, sizeof(T)); + FastMemcpy(&result, ptr, sizeof(T)); } return result; } @@ -758,15 +1006,78 @@ struct Memory::Impl { */ template void Write(Common::ProcessAddress vaddr, const T data) { + // Fast path for aligned writes of common sizes + const u64 addr = GetInteger(vaddr); + if constexpr (std::is_same_v || std::is_same_v) { + // 8-bit writes are always aligned + u8* const ptr = GetPointerImpl( + addr, + [addr, data]() { + LOG_ERROR(HW_Memory, "Unmapped Write8 @ 0x{:016X} = 0x{:02X}", addr, + static_cast(data)); + }, + [&]() { HandleRasterizerWrite(addr, sizeof(T)); }); + if (ptr) { + *ptr = static_cast(data); + } + return; + } else if constexpr (std::is_same_v || std::is_same_v) { + // Check alignment for 16-bit writes + if ((addr & 1) == 0) { + u8* const ptr = GetPointerImpl( + addr, + [addr, data]() { + LOG_ERROR(HW_Memory, "Unmapped Write16 @ 0x{:016X} = 0x{:04X}", addr, + static_cast(data)); + }, + [&]() { HandleRasterizerWrite(addr, sizeof(T)); }); + if (ptr) { + *reinterpret_cast(ptr) = static_cast(data); + return; + } + } + } else if constexpr (std::is_same_v || std::is_same_v) { + // Check alignment for 32-bit writes + if ((addr & 3) == 0) { + u8* const ptr = GetPointerImpl( + addr, + [addr, data]() { + LOG_ERROR(HW_Memory, "Unmapped Write32 @ 0x{:016X} = 0x{:08X}", addr, + static_cast(data)); + }, + [&]() { HandleRasterizerWrite(addr, sizeof(T)); }); + if (ptr) { + *reinterpret_cast(ptr) = static_cast(data); + return; + } + } + } else if constexpr (std::is_same_v || std::is_same_v) { + // Check alignment for 64-bit writes + if ((addr & 7) == 0) { + u8* const ptr = GetPointerImpl( + addr, + [addr, data]() { + LOG_ERROR(HW_Memory, "Unmapped Write64 @ 0x{:016X} = 0x{:016X}", addr, + static_cast(data)); + }, + [&]() { HandleRasterizerWrite(addr, sizeof(T)); }); + if (ptr) { + *reinterpret_cast(ptr) = static_cast(data); + return; + } + } + } + + // Fall back to the general case for other types or unaligned access u8* const ptr = GetPointerImpl( - GetInteger(vaddr), - [vaddr, data]() { + addr, + [addr, data]() { LOG_ERROR(HW_Memory, "Unmapped Write{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8, - GetInteger(vaddr), static_cast(data)); + addr, static_cast(data)); }, - [&]() { HandleRasterizerWrite(GetInteger(vaddr), sizeof(T)); }); + [&]() { HandleRasterizerWrite(addr, sizeof(T)); }); if (ptr) { - std::memcpy(ptr, &data, sizeof(T)); + FastMemcpy(ptr, &data, sizeof(T)); } } @@ -878,6 +1189,12 @@ struct Memory::Impl { Core::System& system; Tegra::MaxwellDeviceMemoryManager* gpu_device_memory{}; Common::PageTable* current_page_table = nullptr; + + // Number of threads to use for parallel memory operations + unsigned int thread_count = 2; + + // Minimum size in bytes for which parallel processing is beneficial + static constexpr size_t PARALLEL_THRESHOLD = 64 * 1024; // 64 KB std::array rasterizer_read_areas{}; std::array rasterizer_write_areas{}; diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.cpp b/src/video_core/renderer_vulkan/renderer_vulkan.cpp index c4fe8235c7..ebc50290a9 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.cpp +++ b/src/video_core/renderer_vulkan/renderer_vulkan.cpp @@ -106,11 +106,20 @@ RendererVulkan::RendererVulkan(Core::Frontend::EmuWindow& emu_window, std::unique_ptr context_) try : RendererBase(emu_window, std::move(context_)), device_memory(device_memory_), gpu(gpu_), library(OpenLibrary(context.get())), + // Create raw Vulkan instance first instance(CreateInstance(*library, dld, VK_API_VERSION_1_1, render_window.GetWindowInfo().type, - Settings::values.renderer_debug.GetValue())), + Settings::values.renderer_debug.GetValue())), + // Now create RAII wrappers for the resources in the correct order + managed_instance(MakeManagedInstance(instance, dld)), + // Create debug messenger if debug is enabled debug_messenger(Settings::values.renderer_debug ? CreateDebugUtilsCallback(instance) - : vk::DebugUtilsMessenger{}), + : vk::DebugUtilsMessenger{}), + managed_debug_messenger(Settings::values.renderer_debug + ? MakeManagedDebugUtilsMessenger(debug_messenger, instance, dld) + : ManagedDebugUtilsMessenger{}), + // Create surface surface(CreateSurface(instance, render_window.GetWindowInfo())), + managed_surface(MakeManagedSurface(surface, instance, dld)), device(CreateDevice(instance, dld, *surface)), memory_allocator(device), state_tracker(), scheduler(device, state_tracker), swapchain(*surface, device, scheduler, render_window.GetFramebufferLayout().width, diff --git a/src/video_core/renderer_vulkan/renderer_vulkan.h b/src/video_core/renderer_vulkan/renderer_vulkan.h index 0a606d6fed..748981cc6f 100644 --- a/src/video_core/renderer_vulkan/renderer_vulkan.h +++ b/src/video_core/renderer_vulkan/renderer_vulkan.h @@ -21,6 +21,7 @@ #include "video_core/vulkan_common/vulkan_memory_allocator.h" #include "video_core/vulkan_common/hybrid_memory.h" #include "video_core/vulkan_common/vulkan_wrapper.h" +#include "video_core/vulkan_common/vulkan_raii.h" namespace Core::Memory { class Memory; @@ -76,9 +77,18 @@ private: std::shared_ptr library; vk::InstanceDispatch dld; + // Keep original handles for compatibility with existing code vk::Instance instance; + // RAII wrapper for instance + ManagedInstance managed_instance; + vk::DebugUtilsMessenger debug_messenger; + // RAII wrapper for debug messenger + ManagedDebugUtilsMessenger managed_debug_messenger; + vk::SurfaceKHR surface; + // RAII wrapper for surface + ManagedSurface managed_surface; Device device; MemoryAllocator memory_allocator; diff --git a/src/video_core/vulkan_common/vulkan_device.cpp b/src/video_core/vulkan_common/vulkan_device.cpp index 9426d31faf..dc7223227a 100644 --- a/src/video_core/vulkan_common/vulkan_device.cpp +++ b/src/video_core/vulkan_common/vulkan_device.cpp @@ -290,9 +290,9 @@ std::unordered_map GetFormatProperties(vk::Physica void OverrideBcnFormats(std::unordered_map& format_properties) { // These properties are extracted from Adreno driver 512.687.0 constexpr VkFormatFeatureFlags tiling_features{ - VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_BLIT_SRC_BIT | - VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT | - VK_FORMAT_FEATURE_TRANSFER_DST_BIT}; + VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_BLIT_SRC_BIT | + VK_FORMAT_FEATURE_SAMPLED_IMAGE_FILTER_LINEAR_BIT | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT | + VK_FORMAT_FEATURE_TRANSFER_DST_BIT}; constexpr VkFormatFeatureFlags buffer_features{VK_FORMAT_FEATURE_UNIFORM_TEXEL_BUFFER_BIT}; @@ -388,11 +388,11 @@ void Device::RemoveExtensionFeature(bool& extension, Feature& feature, // Unload extension. this->RemoveExtension(extension, extension_name); - // Save sType and pNext for chain. + // Save sType and pNext for chain. VkStructureType sType = feature.sType; void* pNext = feature.pNext; - // Clear feature struct and restore chain. + // Clear feature struct and restore chain. feature = {}; feature.sType = sType; feature.pNext = pNext; @@ -442,8 +442,8 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR SetupFamilies(surface); const auto queue_cis = GetDeviceQueueCreateInfos(); - // GetSuitability has already configured the linked list of features for us. - // Reuse it here. + // GetSuitability has already configured the linked list of features for us. + // Reuse it here. const void* first_next = &features2; VkDeviceDiagnosticsConfigCreateInfoNV diagnostics_nv{}; @@ -484,16 +484,10 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR LOG_WARNING(Render_Vulkan, "Qualcomm and Turnip drivers have broken VK_EXT_custom_border_color"); //RemoveExtensionFeature(extensions.custom_border_color, features.custom_border_color, - //VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); + //VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); } if (is_qualcomm) { - must_emulate_scaled_formats = false; - - LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken VK_EXT_extended_dynamic_state"); - //RemoveExtensionFeature(extensions.extended_dynamic_state, features.extended_dynamic_state, - //VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); - LOG_WARNING(Render_Vulkan, "Qualcomm drivers have a slow VK_KHR_push_descriptor implementation"); RemoveExtension(extensions.push_descriptor, VK_KHR_PUSH_DESCRIPTOR_EXTENSION_NAME); @@ -521,14 +515,6 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR #endif } - if (is_arm) { - must_emulate_scaled_formats = false; - - LOG_WARNING(Render_Vulkan, "ARM drivers have broken VK_EXT_extended_dynamic_state"); - //RemoveExtensionFeature(extensions.extended_dynamic_state, features.extended_dynamic_state, - //VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); - } - if (is_nvidia) { const u32 nv_major_version = (properties.properties.driverVersion >> 22) & 0x3ff; const auto arch = GetNvidiaArch(); @@ -553,8 +539,8 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR LOG_WARNING(Render_Vulkan, "RADV versions older than 21.2 have broken VK_EXT_extended_dynamic_state"); //RemoveExtensionFeature(extensions.extended_dynamic_state, - //features.extended_dynamic_state, - //VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); + //features.extended_dynamic_state, + //VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); } } if (extensions.extended_dynamic_state2 && is_radv) { @@ -563,9 +549,9 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR LOG_WARNING( Render_Vulkan, "RADV versions older than 22.3.1 have broken VK_EXT_extended_dynamic_state2"); - // RemoveExtensionFeature(extensions.extended_dynamic_state2, - // features.extended_dynamic_state2, - // VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); + // RemoveExtensionFeature(extensions.extended_dynamic_state2, + // features.extended_dynamic_state2, + // VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); } } if (extensions.extended_dynamic_state2 && is_qualcomm) { @@ -576,8 +562,8 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR LOG_WARNING(Render_Vulkan, "Qualcomm Adreno 7xx drivers have broken VK_EXT_extended_dynamic_state2"); //RemoveExtensionFeature(extensions.extended_dynamic_state2, - //features.extended_dynamic_state2, - //VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); + //features.extended_dynamic_state2, + //VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); } } if (extensions.extended_dynamic_state3 && is_radv) { @@ -610,9 +596,9 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR if (is_rdna2) { LOG_WARNING(Render_Vulkan, "RADV has broken VK_EXT_vertex_input_dynamic_state on RDNA2 hardware"); - // RemoveExtensionFeature(extensions.vertex_input_dynamic_state, - // features.vertex_input_dynamic_state, - // VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); + // RemoveExtensionFeature(extensions.vertex_input_dynamic_state, + // features.vertex_input_dynamic_state, + // VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); } } if (extensions.vertex_input_dynamic_state && is_qualcomm) { @@ -620,8 +606,8 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR LOG_WARNING(Render_Vulkan, "Qualcomm drivers have broken VK_EXT_vertex_input_dynamic_state"); //RemoveExtensionFeature(extensions.vertex_input_dynamic_state, - // features.vertex_input_dynamic_state, - // VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); + // features.vertex_input_dynamic_state, + // VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); } sets_per_pool = 64; @@ -725,6 +711,37 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR dynamic_state3_enables = true; } + if (Settings::values.dyna_state.GetValue() == 0) { + must_emulate_scaled_formats = true; + LOG_INFO(Render_Vulkan, "Dynamic state is disabled (dyna_state = 0), forcing scaled format emulation ON"); + + // Remove all dynamic state 1-2 extensions and features + RemoveExtensionFeature(extensions.custom_border_color, features.custom_border_color, + VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); + + RemoveExtensionFeature(extensions.extended_dynamic_state, features.extended_dynamic_state, + VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); + + RemoveExtensionFeature(extensions.extended_dynamic_state2, features.extended_dynamic_state2, + VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); + + RemoveExtensionFeature(extensions.vertex_input_dynamic_state, features.vertex_input_dynamic_state, + VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); + + // Disable extended dynamic state 3 features + features.extended_dynamic_state3.extendedDynamicState3ColorBlendEnable = false; + features.extended_dynamic_state3.extendedDynamicState3ColorBlendEquation = false; + features.extended_dynamic_state3.extendedDynamicState3DepthClampEnable = false; + + dynamic_state3_blending = false; + dynamic_state3_enables = false; + + LOG_INFO(Render_Vulkan, "Dynamic state extensions and features have been fully disabled"); + } else { + must_emulate_scaled_formats = false; + LOG_INFO(Render_Vulkan, "Dynamic state is enabled (dyna_state = 1-3), disabling scaled format emulation"); + } + logical = vk::Device::Create(physical, queue_cis, ExtensionListForVulkan(loaded_extensions), first_next, dld); @@ -782,7 +799,7 @@ VkFormat Device::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags return alternative; } - // No alternatives found, panic + // No alternatives found, panic LOG_ERROR(Render_Vulkan, "Format={} with usage={} and type={} is not supported by the host hardware and " "doesn't support any of the alternatives", @@ -793,7 +810,7 @@ VkFormat Device::GetSupportedFormat(VkFormat wanted_format, VkFormatFeatureFlags void Device::ReportLoss() const { LOG_CRITICAL(Render_Vulkan, "Device loss occurred!"); - // Wait for the log to flush and for Nsight Aftermath to dump the results + // Wait for the log to flush and for Nsight Aftermath to dump the results std::this_thread::sleep_for(std::chrono::seconds{15}); } @@ -825,9 +842,9 @@ bool Device::ComputeIsOptimalAstcSupported() const { return false; } const auto format_feature_usage{ - VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_BLIT_SRC_BIT | - VK_FORMAT_FEATURE_BLIT_DST_BIT | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT | - VK_FORMAT_FEATURE_TRANSFER_DST_BIT}; + VK_FORMAT_FEATURE_SAMPLED_IMAGE_BIT | VK_FORMAT_FEATURE_BLIT_SRC_BIT | + VK_FORMAT_FEATURE_BLIT_DST_BIT | VK_FORMAT_FEATURE_TRANSFER_SRC_BIT | + VK_FORMAT_FEATURE_TRANSFER_DST_BIT}; for (const auto format : astc_formats) { const auto physical_format_properties{physical.GetFormatProperties(format)}; if ((physical_format_properties.optimalTilingFeatures & format_feature_usage) == 0) { @@ -946,36 +963,36 @@ bool Device::GetSuitability(bool requires_swapchain) { // Assume we will be suitable. bool suitable = true; - // Configure properties. + // Configure properties. properties.properties = physical.GetProperties(); - // Set instance version. + // Set instance version. instance_version = properties.properties.apiVersion; - // Minimum of API version 1.1 is required. (This is well-supported.) + // Minimum of API version 1.1 is required. (This is well-supported.) ASSERT(instance_version >= VK_API_VERSION_1_1); - // Get available extensions. + // Get available extensions. auto extension_properties = physical.EnumerateDeviceExtensionProperties(); - // Get the set of supported extensions. + // Get the set of supported extensions. supported_extensions.clear(); for (const VkExtensionProperties& property : extension_properties) { supported_extensions.insert(property.extensionName); } - // Generate list of extensions to load. + // Generate list of extensions to load. loaded_extensions.clear(); #define EXTENSION(prefix, macro_name, var_name) \ if (supported_extensions.contains(VK_##prefix##_##macro_name##_EXTENSION_NAME)) { \ - loaded_extensions.insert(VK_##prefix##_##macro_name##_EXTENSION_NAME); \ - extensions.var_name = true; \ + loaded_extensions.insert(VK_##prefix##_##macro_name##_EXTENSION_NAME); \ + extensions.var_name = true; \ } #define FEATURE_EXTENSION(prefix, struct_name, macro_name, var_name) \ if (supported_extensions.contains(VK_##prefix##_##macro_name##_EXTENSION_NAME)) { \ - loaded_extensions.insert(VK_##prefix##_##macro_name##_EXTENSION_NAME); \ - extensions.var_name = true; \ + loaded_extensions.insert(VK_##prefix##_##macro_name##_EXTENSION_NAME); \ + extensions.var_name = true; \ } if (instance_version < VK_API_VERSION_1_2) { @@ -991,16 +1008,16 @@ bool Device::GetSuitability(bool requires_swapchain) { #undef FEATURE_EXTENSION #undef EXTENSION - // Some extensions are mandatory. Check those. + // Some extensions are mandatory. Check those. #define CHECK_EXTENSION(extension_name) \ if (!loaded_extensions.contains(extension_name)) { \ - LOG_ERROR(Render_Vulkan, "Missing required extension {}", extension_name); \ - suitable = false; \ + LOG_ERROR(Render_Vulkan, "Missing required extension {}", extension_name); \ + suitable = false; \ } #define LOG_EXTENSION(extension_name) \ if (!loaded_extensions.contains(extension_name)) { \ - LOG_INFO(Render_Vulkan, "Device doesn't support extension {}", extension_name); \ + LOG_INFO(Render_Vulkan, "Device doesn't support extension {}", extension_name); \ } FOR_EACH_VK_RECOMMENDED_EXTENSION(LOG_EXTENSION); @@ -1013,24 +1030,24 @@ bool Device::GetSuitability(bool requires_swapchain) { #undef LOG_EXTENSION #undef CHECK_EXTENSION - // Generate the linked list of features to test. + // Generate the linked list of features to test. features2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2; - // Set next pointer. + // Set next pointer. void** next = &features2.pNext; - // Test all features we know about. If the feature is not available in core at our - // current API version, and was not enabled by an extension, skip testing the feature. - // We set the structure sType explicitly here as it is zeroed by the constructor. + // Test all features we know about. If the feature is not available in core at our + // current API version, and was not enabled by an extension, skip testing the feature. + // We set the structure sType explicitly here as it is zeroed by the constructor. #define FEATURE(prefix, struct_name, macro_name, var_name) \ features.var_name.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_##macro_name##_FEATURES; \ - SetNext(next, features.var_name); + SetNext(next, features.var_name); #define EXT_FEATURE(prefix, struct_name, macro_name, var_name) \ if (extensions.var_name) { \ - features.var_name.sType = \ + features.var_name.sType = \ VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_##macro_name##_FEATURES_##prefix; \ - SetNext(next, features.var_name); \ + SetNext(next, features.var_name); \ } FOR_EACH_VK_FEATURE_1_1(FEATURE); @@ -1049,20 +1066,20 @@ bool Device::GetSuitability(bool requires_swapchain) { #undef EXT_FEATURE #undef FEATURE - // Perform the feature test. + // Perform the feature test. physical.GetFeatures2(features2); features.features = features2.features; - // Some features are mandatory. Check those. + // Some features are mandatory. Check those. #define CHECK_FEATURE(feature, name) \ if (!features.feature.name) { \ - LOG_ERROR(Render_Vulkan, "Missing required feature {}", #name); \ - suitable = false; \ + LOG_ERROR(Render_Vulkan, "Missing required feature {}", #name); \ + suitable = false; \ } #define LOG_FEATURE(feature, name) \ if (!features.feature.name) { \ - LOG_INFO(Render_Vulkan, "Device doesn't support feature {}", #name); \ + LOG_INFO(Render_Vulkan, "Device doesn't support feature {}", #name); \ } FOR_EACH_VK_RECOMMENDED_FEATURE(LOG_FEATURE); @@ -1071,21 +1088,21 @@ bool Device::GetSuitability(bool requires_swapchain) { #undef LOG_FEATURE #undef CHECK_FEATURE - // Generate linked list of properties. + // Generate linked list of properties. properties2.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROPERTIES_2; - // Set next pointer. + // Set next pointer. next = &properties2.pNext; - // Get driver info. + // Get driver info. properties.driver.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES; SetNext(next, properties.driver); - // Retrieve subgroup properties. + // Retrieve subgroup properties. properties.subgroup_properties.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_PROPERTIES; SetNext(next, properties.subgroup_properties); - // Retrieve relevant extension properties. + // Retrieve relevant extension properties. if (extensions.shader_float_controls) { properties.float_controls.sType = VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FLOAT_CONTROLS_PROPERTIES; @@ -1107,14 +1124,14 @@ bool Device::GetSuitability(bool requires_swapchain) { SetNext(next, properties.transform_feedback); } - // Perform the property fetch. + // Perform the property fetch. physical.GetProperties2(properties2); properties.properties = properties2.properties; - // Unload extensions if feature support is insufficient. + // Unload extensions if feature support is insufficient. RemoveUnsuitableExtensions(); - // Check limits. + // Check limits. struct Limit { u32 minimum; u32 value; @@ -1123,11 +1140,11 @@ bool Device::GetSuitability(bool requires_swapchain) { const VkPhysicalDeviceLimits& limits{properties.properties.limits}; const std::array limits_report{ - Limit{65536, limits.maxUniformBufferRange, "maxUniformBufferRange"}, - Limit{16, limits.maxViewports, "maxViewports"}, - Limit{8, limits.maxColorAttachments, "maxColorAttachments"}, - Limit{8, limits.maxClipDistances, "maxClipDistances"}, - }; + Limit{65536, limits.maxUniformBufferRange, "maxUniformBufferRange"}, + Limit{16, limits.maxViewports, "maxViewports"}, + Limit{8, limits.maxColorAttachments, "maxColorAttachments"}, + Limit{8, limits.maxClipDistances, "maxClipDistances"}, + }; for (const auto& [min, value, name] : limits_report) { if (value < min) { @@ -1136,7 +1153,7 @@ bool Device::GetSuitability(bool requires_swapchain) { } } - // Return whether we were suitable. + // Return whether we were suitable. return suitable; } @@ -1147,31 +1164,31 @@ void Device::RemoveUnsuitableExtensions() { RemoveExtensionFeatureIfUnsuitable(extensions.custom_border_color, features.custom_border_color, VK_EXT_CUSTOM_BORDER_COLOR_EXTENSION_NAME); - // VK_EXT_depth_bias_control + // VK_EXT_depth_bias_control extensions.depth_bias_control = features.depth_bias_control.depthBiasControl && features.depth_bias_control.leastRepresentableValueForceUnormRepresentation; RemoveExtensionFeatureIfUnsuitable(extensions.depth_bias_control, features.depth_bias_control, VK_EXT_DEPTH_BIAS_CONTROL_EXTENSION_NAME); - // VK_EXT_depth_clip_control + // VK_EXT_depth_clip_control extensions.depth_clip_control = features.depth_clip_control.depthClipControl; RemoveExtensionFeatureIfUnsuitable(extensions.depth_clip_control, features.depth_clip_control, VK_EXT_DEPTH_CLIP_CONTROL_EXTENSION_NAME); - /* */ // VK_EXT_extended_dynamic_state + /* */ // VK_EXT_extended_dynamic_state extensions.extended_dynamic_state = features.extended_dynamic_state.extendedDynamicState; RemoveExtensionFeatureIfUnsuitable(extensions.extended_dynamic_state, features.extended_dynamic_state, VK_EXT_EXTENDED_DYNAMIC_STATE_EXTENSION_NAME); - // VK_EXT_extended_dynamic_state2 + // VK_EXT_extended_dynamic_state2 extensions.extended_dynamic_state2 = features.extended_dynamic_state2.extendedDynamicState2; RemoveExtensionFeatureIfUnsuitable(extensions.extended_dynamic_state2, features.extended_dynamic_state2, VK_EXT_EXTENDED_DYNAMIC_STATE_2_EXTENSION_NAME); - // VK_EXT_extended_dynamic_state3 + // VK_EXT_extended_dynamic_state3 dynamic_state3_blending = features.extended_dynamic_state3.extendedDynamicState3ColorBlendEnable && features.extended_dynamic_state3.extendedDynamicState3ColorBlendEquation && @@ -1187,27 +1204,27 @@ void Device::RemoveUnsuitableExtensions() { features.extended_dynamic_state3, VK_EXT_EXTENDED_DYNAMIC_STATE_3_EXTENSION_NAME); - // VK_EXT_provoking_vertex + // VK_EXT_provoking_vertex extensions.provoking_vertex = features.provoking_vertex.provokingVertexLast && features.provoking_vertex.transformFeedbackPreservesProvokingVertex; RemoveExtensionFeatureIfUnsuitable(extensions.provoking_vertex, features.provoking_vertex, VK_EXT_PROVOKING_VERTEX_EXTENSION_NAME); - // VK_KHR_shader_atomic_int64 + // VK_KHR_shader_atomic_int64 extensions.shader_atomic_int64 = features.shader_atomic_int64.shaderBufferInt64Atomics && features.shader_atomic_int64.shaderSharedInt64Atomics; RemoveExtensionFeatureIfUnsuitable(extensions.shader_atomic_int64, features.shader_atomic_int64, VK_KHR_SHADER_ATOMIC_INT64_EXTENSION_NAME); - // VK_EXT_shader_demote_to_helper_invocation + // VK_EXT_shader_demote_to_helper_invocation extensions.shader_demote_to_helper_invocation = features.shader_demote_to_helper_invocation.shaderDemoteToHelperInvocation; RemoveExtensionFeatureIfUnsuitable(extensions.shader_demote_to_helper_invocation, features.shader_demote_to_helper_invocation, VK_EXT_SHADER_DEMOTE_TO_HELPER_INVOCATION_EXTENSION_NAME); - // VK_EXT_subgroup_size_control + // VK_EXT_subgroup_size_control extensions.subgroup_size_control = features.subgroup_size_control.subgroupSizeControl && properties.subgroup_size_control.minSubgroupSize <= GuestWarpSize && @@ -1216,7 +1233,7 @@ void Device::RemoveUnsuitableExtensions() { features.subgroup_size_control, VK_EXT_SUBGROUP_SIZE_CONTROL_EXTENSION_NAME); - // VK_EXT_transform_feedback + // VK_EXT_transform_feedback extensions.transform_feedback = features.transform_feedback.transformFeedback && features.transform_feedback.geometryStreams && @@ -1227,14 +1244,14 @@ void Device::RemoveUnsuitableExtensions() { RemoveExtensionFeatureIfUnsuitable(extensions.transform_feedback, features.transform_feedback, VK_EXT_TRANSFORM_FEEDBACK_EXTENSION_NAME); - // VK_EXT_vertex_input_dynamic_state + // VK_EXT_vertex_input_dynamic_state extensions.vertex_input_dynamic_state = features.vertex_input_dynamic_state.vertexInputDynamicState; RemoveExtensionFeatureIfUnsuitable(extensions.vertex_input_dynamic_state, features.vertex_input_dynamic_state, VK_EXT_VERTEX_INPUT_DYNAMIC_STATE_EXTENSION_NAME); - // VK_KHR_pipeline_executable_properties + // VK_KHR_pipeline_executable_properties if (Settings::values.renderer_shader_feedback.GetValue()) { extensions.pipeline_executable_properties = features.pipeline_executable_properties.pipelineExecutableInfo; @@ -1247,7 +1264,7 @@ void Device::RemoveUnsuitableExtensions() { VK_KHR_PIPELINE_EXECUTABLE_PROPERTIES_EXTENSION_NAME); } - // VK_KHR_workgroup_memory_explicit_layout + // VK_KHR_workgroup_memory_explicit_layout extensions.workgroup_memory_explicit_layout = features.features.shaderInt16 && features.workgroup_memory_explicit_layout.workgroupMemoryExplicitLayout && diff --git a/src/video_core/vulkan_common/vulkan_raii.h b/src/video_core/vulkan_common/vulkan_raii.h new file mode 100644 index 0000000000..cf5e268b68 --- /dev/null +++ b/src/video_core/vulkan_common/vulkan_raii.h @@ -0,0 +1,231 @@ +// SPDX-FileCopyrightText: Copyright 2025 yuzu Emulator Project +// SPDX-License-Identifier: GPL-2.0-or-later + +#pragma once + +#include +#include +#include +#include + +#include "common/logging/log.h" + +#include "video_core/vulkan_common/vulkan_wrapper.h" + +namespace Vulkan { + +/** + * RAII wrapper for Vulkan resources. + * Automatically manages the lifetime of Vulkan objects using RAII principles. + */ +template +class VulkanRaii { +public: + using DeleterFunc = std::function; + + // Default constructor - creates a null handle + VulkanRaii() : handle{}, deleter{}, dispatch{} {} + + // Constructor with handle and deleter + VulkanRaii(T handle_, DeleterFunc deleter_, const Dispatch& dispatch_, const char* resource_name = "Vulkan resource") + : handle{handle_}, deleter{std::move(deleter_)}, dispatch{dispatch_} { + LOG_DEBUG(Render_Vulkan, "RAII wrapper created for {}", resource_name); + } + + // Move constructor + VulkanRaii(VulkanRaii&& other) noexcept + : handle{std::exchange(other.handle, VK_NULL_HANDLE)}, + deleter{std::move(other.deleter)}, + dispatch{other.dispatch} { + } + + // Move assignment + VulkanRaii& operator=(VulkanRaii&& other) noexcept { + if (this != &other) { + cleanup(); + handle = std::exchange(other.handle, VK_NULL_HANDLE); + deleter = std::move(other.deleter); + dispatch = other.dispatch; + } + return *this; + } + + // Destructor - automatically cleans up the resource + ~VulkanRaii() { + cleanup(); + } + + // Disallow copying + VulkanRaii(const VulkanRaii&) = delete; + VulkanRaii& operator=(const VulkanRaii&) = delete; + + // Get the underlying handle + T get() const noexcept { + return handle; + } + + // Check if the handle is valid + bool valid() const noexcept { + return handle != VK_NULL_HANDLE; + } + + // Release ownership of the handle without destroying it + T release() noexcept { + return std::exchange(handle, VK_NULL_HANDLE); + } + + // Reset the handle (destroying the current one if it exists) + void reset(T new_handle = VK_NULL_HANDLE, DeleterFunc new_deleter = {}) { + cleanup(); + handle = new_handle; + deleter = std::move(new_deleter); + } + + // Implicit conversion to handle type + operator T() const noexcept { + return handle; + } + + // Dereference operator for pointer-like access + T operator->() const noexcept { + return handle; + } + +private: + // Optimized cleanup function + void cleanup() noexcept { + if (handle != VK_NULL_HANDLE && deleter) { + deleter(handle, dispatch); + handle = VK_NULL_HANDLE; + } + } + + T handle; + DeleterFunc deleter; + Dispatch dispatch; +}; + +// Common type aliases for Vulkan RAII wrappers with clearer names +using ManagedInstance = VulkanRaii; +using ManagedDevice = VulkanRaii; +using ManagedSurface = VulkanRaii; +using ManagedSwapchain = VulkanRaii; +using ManagedCommandPool = VulkanRaii; +using ManagedBuffer = VulkanRaii; +using ManagedImage = VulkanRaii; +using ManagedImageView = VulkanRaii; +using ManagedSampler = VulkanRaii; +using ManagedShaderModule = VulkanRaii; +using ManagedPipeline = VulkanRaii; +using ManagedPipelineLayout = VulkanRaii; +using ManagedDescriptorSetLayout = VulkanRaii; +using ManagedDescriptorPool = VulkanRaii; +using ManagedSemaphore = VulkanRaii; +using ManagedFence = VulkanRaii; +using ManagedDebugUtilsMessenger = VulkanRaii; + +// Helper functions to create RAII wrappers + +/** + * Creates an RAII wrapper for a Vulkan instance + */ +inline ManagedInstance MakeManagedInstance(const vk::Instance& instance, const vk::InstanceDispatch& dispatch) { + auto deleter = [](VkInstance handle, const vk::InstanceDispatch& dld) { + dld.vkDestroyInstance(handle, nullptr); + }; + return ManagedInstance(*instance, deleter, dispatch, "VkInstance"); +} + +/** + * Creates an RAII wrapper for a Vulkan device + */ +inline ManagedDevice MakeManagedDevice(const vk::Device& device, const vk::DeviceDispatch& dispatch) { + auto deleter = [](VkDevice handle, const vk::DeviceDispatch& dld) { + dld.vkDestroyDevice(handle, nullptr); + }; + return ManagedDevice(*device, deleter, dispatch, "VkDevice"); +} + +/** + * Creates an RAII wrapper for a Vulkan surface + */ +inline ManagedSurface MakeManagedSurface(const vk::SurfaceKHR& surface, const vk::Instance& instance, const vk::InstanceDispatch& dispatch) { + auto deleter = [instance_ptr = *instance](VkSurfaceKHR handle, const vk::InstanceDispatch& dld) { + dld.vkDestroySurfaceKHR(instance_ptr, handle, nullptr); + }; + return ManagedSurface(*surface, deleter, dispatch, "VkSurfaceKHR"); +} + +/** + * Creates an RAII wrapper for a Vulkan debug messenger + */ +inline ManagedDebugUtilsMessenger MakeManagedDebugUtilsMessenger(const vk::DebugUtilsMessenger& messenger, + const vk::Instance& instance, + const vk::InstanceDispatch& dispatch) { + auto deleter = [instance_ptr = *instance](VkDebugUtilsMessengerEXT handle, const vk::InstanceDispatch& dld) { + dld.vkDestroyDebugUtilsMessengerEXT(instance_ptr, handle, nullptr); + }; + return ManagedDebugUtilsMessenger(*messenger, deleter, dispatch, "VkDebugUtilsMessengerEXT"); +} + +/** + * Creates an RAII wrapper for a Vulkan swapchain + */ +inline ManagedSwapchain MakeManagedSwapchain(VkSwapchainKHR swapchain_handle, VkDevice device_handle, const vk::DeviceDispatch& dispatch) { + auto deleter = [device_handle](VkSwapchainKHR handle, const vk::DeviceDispatch& dld) { + dld.vkDestroySwapchainKHR(device_handle, handle, nullptr); + }; + return ManagedSwapchain(swapchain_handle, deleter, dispatch, "VkSwapchainKHR"); +} + +/** + * Creates an RAII wrapper for a Vulkan buffer + */ +inline ManagedBuffer MakeManagedBuffer(VkBuffer buffer_handle, VkDevice device_handle, const vk::DeviceDispatch& dispatch) { + auto deleter = [device_handle](VkBuffer handle, const vk::DeviceDispatch& dld) { + dld.vkDestroyBuffer(device_handle, handle, nullptr); + }; + return ManagedBuffer(buffer_handle, deleter, dispatch, "VkBuffer"); +} + +/** + * Creates an RAII wrapper for a Vulkan image + */ +inline ManagedImage MakeManagedImage(VkImage image_handle, VkDevice device_handle, const vk::DeviceDispatch& dispatch) { + auto deleter = [device_handle](VkImage handle, const vk::DeviceDispatch& dld) { + dld.vkDestroyImage(device_handle, handle, nullptr); + }; + return ManagedImage(image_handle, deleter, dispatch, "VkImage"); +} + +/** + * Creates an RAII wrapper for a Vulkan image view + */ +inline ManagedImageView MakeManagedImageView(VkImageView view_handle, VkDevice device_handle, const vk::DeviceDispatch& dispatch) { + auto deleter = [device_handle](VkImageView handle, const vk::DeviceDispatch& dld) { + dld.vkDestroyImageView(device_handle, handle, nullptr); + }; + return ManagedImageView(view_handle, deleter, dispatch, "VkImageView"); +} + +/** + * Creates an RAII wrapper for a Vulkan semaphore + */ +inline ManagedSemaphore MakeManagedSemaphore(VkSemaphore semaphore_handle, VkDevice device_handle, const vk::DeviceDispatch& dispatch) { + auto deleter = [device_handle](VkSemaphore handle, const vk::DeviceDispatch& dld) { + dld.vkDestroySemaphore(device_handle, handle, nullptr); + }; + return ManagedSemaphore(semaphore_handle, deleter, dispatch, "VkSemaphore"); +} + +/** + * Creates an RAII wrapper for a Vulkan fence + */ +inline ManagedFence MakeManagedFence(VkFence fence_handle, VkDevice device_handle, const vk::DeviceDispatch& dispatch) { + auto deleter = [device_handle](VkFence handle, const vk::DeviceDispatch& dld) { + dld.vkDestroyFence(device_handle, handle, nullptr); + }; + return ManagedFence(fence_handle, deleter, dispatch, "VkFence"); +} + +} // namespace Vulkan \ No newline at end of file